Infrastructure for continually self‑improving agents
This report summarizes scanner evidence for Human-Agent-Society/reef, a Python project with 7,521 GitHub stars. SynthScan examined 289,501 lines of code across 1615 source files, recording 6688 pattern matches distributed across 21 syntactic categories. The legacy adjusted pattern signal is 27.4 (Elevated pattern signal); it is not an authorship estimate.
Use the findings to ask whether maintainers can explain intent, verify dependencies, simplify unnecessary scaffolding, and focus review effort. Risk Payload v1 adds ownership, supply-chain, over-engineering, and review-effort vectors when available. These are prompts for human verification, not probabilities, security findings, or measured maintenance costs.
Use each vector to guide manual verification of the underlying findings.
Can maintainers explain intent, assumptions, and trade-offs? 4906 finding(s).
Are imports declared and package/API names verified? 1234 finding(s).
Are wrappers, guards, and repeated structures necessary? 4955 finding(s).
Which changes need focused human review? 6688 finding(s).
Run additional scans to build a history of the legacy pattern signal. Score changes are not a measure of AI adoption or code quality.
Groups findings by configured severity weight. Severity is not a confidence estimate, security rating, or proof that a package is missing. Verify imports against the project's manifests and inspect each finding in its local design context.
This horizontal bar chart shows where configured pattern scores accumulate across directories. Treat the bars as locations to inspect, not as evidence of authorship or code quality. Use the file-level findings to decide whether a directory needs a focused ownership, dependency, complexity, or review check.
The scanner identified 6688 distinct pattern matches across 21 syntactic categories. Each entry below represents a discrete location in the source code where the engine recorded a detector match to inspect in context. Expand any category row to inspect the individual file paths, line numbers, code snippets, and the lexical context (CODE, COMMENT, or STRING) in which each match was detected.
Reading the findings table: The Severity column indicates the configured finding weight (CRITICAL / HIGH / MEDIUM / LOW). The Context column identifies whether the match occurred inside executable code, an inline comment, or a string literal — legacy comment-context matches receive a ×1.5 score weight. The ⚡ bolt icon marks clustered matches: three or more patterns within a 10-line window, each receiving an additional ×1.5 density multiplier under the legacy scoring model.
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW⚡ | tests/test_sdpo_sciknoweval.py | 48 | def test_the_split_loads_as_the_reference_writes_it(chemistry) -> None: | CODE |
| LOW⚡ | tests/test_sdpo_sciknoweval.py | 57 | def test_the_request_is_the_reference_system_prompt_then_the_question(chemistry) -> None: | CODE |
| LOW⚡ | tests/test_sdpo_sciknoweval.py | 65 | def test_the_answer_rule_follows_the_reference_scorer(chemistry) -> None: | CODE |
| LOW | tests/test_sdpo_sciknoweval.py | 78 | def test_the_question_schedule_shuffles_each_epoch_and_drops_only_the_tail(chemistry) -> None: | CODE |
| LOW | tests/test_sdpo_sciknoweval.py | 94 | def test_the_grid_is_sampled_in_one_pool_and_kept_in_group_and_rollout_order(chemistry) -> None: | CODE |
| LOW | tests/test_sdpo_sciknoweval.py | 113 | def test_evaluation_is_unreported_and_asks_one_choice_per_request(chemistry) -> None: | CODE |
| LOW⚡ | tests/test_meta_harness_terminal_bench.py | 62 | def test_binary_verifier_scores(driver, score): | CODE |
| LOW⚡ | tests/test_meta_harness_terminal_bench.py | 67 | def test_invalid_verifier_rewards_refuse(driver, score): | CODE |
| LOW⚡ | tests/test_meta_harness_terminal_bench.py | 72 | def test_missing_failed_and_mismatched_verifier(driver): | CODE |
| LOW⚡ | tests/test_meta_harness_terminal_bench.py | 79 | def test_task_pin_and_drift_checks(driver, tmp_path, monkeypatch): | CODE |
| LOW | tests/test_meta_harness_terminal_bench.py | 108 | def test_configuration_scales_budgets_and_isolates_campaigns(driver, monkeypatch): | CODE |
| LOW | tests/test_meta_harness_terminal_bench.py | 154 | def test_search_commits_and_restart_repairs_mirrors_without_more_episodes(driver, campaign): | CODE |
| LOW | tests/test_meta_harness_terminal_bench.py | 182 | def test_commit_failure_writes_no_summary_and_restart_replays_pending_report(driver, campaign, monkeypatch): | CODE |
| LOW | tests/test_meta_harness_terminal_bench.py | 211 | def test_rollout_record_survives_interrupted_report(driver, campaign, monkeypatch): | CODE |
| LOW | tests/test_coral_bundle.py | 43 | def test_bundle_scores_best_revisions_and_tokens(tmp_path): | CODE |
| LOW | tests/test_coral_bundle.py | 76 | def test_bundle_without_references_falls_back_to_the_parent_coordinate(tmp_path): | CODE |
| LOW | tests/test_coral_bundle.py | 91 | def test_bundle_filters_by_run_and_tolerates_unscored(tmp_path): | CODE |
| LOW | tests/test_coral_bundle.py | 102 | def test_middleware_captures_release_and_usage(tmp_path): | CODE |
| LOW | tests/test_coral_bundle.py | 153 | def test_middleware_usage_from_final_sse_chunk(tmp_path): | CODE |
| LOW | tests/test_tttd_judge.py | 30 | def test_judge_scores_independent_submissions_concurrently(monkeypatch, tmp_path, server_path): | CODE |
| LOW | tests/test_spade_generation.py | 28 | def test_the_experience_is_sorted_into_frontier_mastered_and_out_of_reach_with_the_frontier_by_regret() -> None: | CODE |
| LOW | tests/test_spade_generation.py | 51 | def test_only_frontier_records_show_their_instruction_and_a_long_excerpt_is_cut() -> None: | CODE |
| LOW | tests/test_spade_generation.py | 77 | def test_a_play_record_knows_its_outcome_and_regret( | CODE |
| LOW | tests/test_spade_generation.py | 94 | def test_a_bad_play_record_is_refused(fields: dict[str, object], message: str) -> None: | CODE |
| LOW | tests/test_spade_generation.py | 106 | def test_experience_for_puts_this_skill_and_the_frontier_first_and_caps_the_records() -> None: | CODE |
| LOW | tests/test_spade_generation.py | 132 | def test_mean_reward_counts_an_unscored_run_as_a_loss_and_skips_an_episode_that_never_ran(tmp_path: Path) -> None: | CODE |
| LOW | tests/test_spade_generation.py | 141 | def test_a_generation_report_keeps_the_proposals_the_measures_and_the_experience(tmp_path: Path) -> None: | CODE |
| LOW | tests/test_spade_generation.py | 168 | def test_load_experience_refuses_a_file_that_is_not_a_report(tmp_path: Path) -> None: | CODE |
| LOW | tests/test_tttd_harness.py | 27 | def test_puct_records_top_children_and_backpropagates_visits(): | CODE |
| LOW | tests/test_tttd_harness.py | 43 | def test_puct_archive_snapshot_round_trip_preserves_search_state(): | CODE |
| LOW | tests/test_tttd_harness.py | 64 | def test_puct_pruning_always_retains_seeds(): | CODE |
| LOW | tests/test_tttd_harness.py | 74 | def test_parent_prompt_reuses_the_normalized_code_block_without_nesting(): | CODE |
| LOW⚡ | tests/test_tttd_harness.py | 86 | def test_solution_extraction_matches_official_unclosed_final_block_behavior(): | CODE |
| LOW⚡ | tests/test_tttd_harness.py | 93 | def test_packing_instruction_does_not_claim_the_dynamic_parent_is_empty(task): | CODE |
| LOW⚡ | tests/test_tttd_harness.py | 100 | def test_erdos_instruction_and_judge_use_the_same_program_budget(): | CODE |
| LOW | tests/test_tttd_harness.py | 119 | def test_erdos_scorer_scores_valid_construction(): | CODE |
| LOW | tests/test_tttd_harness.py | 141 | def test_erdos_scorer_rejects_invalid_solution(): | CODE |
| LOW | tests/test_tttd_harness.py | 174 | def test_ordinary_harness_calls_model_evaluates_and_archives(): | CODE |
| LOW | tests/test_tttd_harness.py | 216 | def test_reef_harness_reports_each_result_against_exact_inference(): | CODE |
| LOW | tests/test_tttd_harness.py | 245 | def test_tttd_chat_request_builder_leaves_token_capture_to_backend(): | CODE |
| LOW⚡ | tests/test_tttd_harness.py | 261 | def test_tttd_chat_request_builder_rejects_adapter_override(): | CODE |
| LOW⚡ | tests/test_tttd_harness.py | 268 | def test_sandbox_environment_excludes_operator_secrets(monkeypatch): | CODE |
| LOW | tests/test_coral_gateway_launcher.py | 47 | def test_attach_splices_under_coral_and_keeps_register_agent_working(tmp_path): | CODE |
| LOW | tests/test_coral_gateway_launcher.py | 63 | def test_insert_is_idempotent(tmp_path): | CODE |
| LOW | tests/test_coral_gateway_launcher.py | 74 | def test_insert_requires_a_started_middleware(tmp_path): | CODE |
| LOW | tests/test_coral_gateway_launcher.py | 90 | def _start_gateway_if_enabled(self): | CODE |
| LOW | tests/test_coral_gateway_launcher.py | 101 | def test_attach_to_agent_manager_splices_before_agents_spawn(tmp_path): | CODE |
| LOW | tests/test_coral_gateway_launcher.py | 112 | def test_attach_to_agent_manager_fails_loudly_when_gateway_disabled(tmp_path): | CODE |
| LOW | tests/test_example_startup.py | 28 | def test_example_startup_and_cleanup(tmp_path: Path, example: str, mode: str) -> None: | CODE |
| LOW | tests/test_reef_client_serve.py | 44 | def test_sse_accumulator_round_trip_with_synthesized_events() -> None: | CODE |
| LOW | tests/test_reef_client_serve.py | 63 | def test_sse_accumulator_ignores_keepalives_and_partial_events() -> None: | CODE |
| LOW | tests/test_reef_client_serve.py | 76 | def test_sse_accumulator_accumulates_tool_call_argument_fragments() -> None: | CODE |
| LOW | tests/test_reef_client_serve.py | 188 | def test_serve_stamps_session_and_captures_buffered_turn() -> None: | CODE |
| LOW | tests/test_reef_client_serve.py | 223 | def test_serve_relays_sse_passthrough_and_reconstructs_capture() -> None: | CODE |
| LOW | tests/test_reef_client_serve.py | 258 | def test_serve_force_non_stream_synthesizes_sse_for_streaming_clients() -> None: | CODE |
| LOW | tests/test_reef_client_serve.py | 296 | def test_serve_reports_upstream_error_in_committed_stream() -> None: | CODE |
| LOW | tests/test_reef_client_serve.py | 329 | def test_serve_captures_endpoint_snapshots_and_clears() -> None: | CODE |
| LOW | tests/test_reef_client_serve.py | 366 | def test_serve_addressable_sessions_self_identify_by_path() -> None: | CODE |
| LOW | tests/test_reef_client_serve.py | 406 | def test_serve_addressable_sessions_coexist_with_legacy_global_mode() -> None: | CODE |
| LOW | tests/test_reef_client_serve.py | 424 | def test_serve_session_lifecycle_and_validation() -> None: | CODE |
| 4328 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tests/test_sdpo_sciknoweval.py | 8 | CODE | |
| LOW | tests/conftest.py | 13 | CODE | |
| LOW | tests/test_meta_harness_terminal_bench.py | 3 | CODE | |
| LOW | tests/test_coral_bundle.py | 3 | CODE | |
| LOW | tests/test_tttd_judge.py | 1 | CODE | |
| LOW | tests/test_spade_generation.py | 3 | CODE | |
| LOW | tests/test_tttd_harness.py | 1 | CODE | |
| LOW | tests/test_coral_gateway_launcher.py | 3 | CODE | |
| LOW | tests/test_reef_client_serve.py | 3 | CODE | |
| LOW | tests/test_coral_watcher.py | 9 | CODE | |
| LOW | tests/test_openclawrl.py | 3 | CODE | |
| LOW | tests/test_spade_training.py | 3 | CODE | |
| LOW | tests/test_example_entrypoints.py | 3 | CODE | |
| LOW | tests/test_coral_body_headers.py | 3 | CODE | |
| LOW | tests/test_sdft_skill_stream.py | 3 | CODE | |
| LOW | tests/test_guidance_ttt.py | 1 | CODE | |
| LOW | tests/test_openclawrl_stream.py | 3 | CODE | |
| LOW | tests/test_gepa_aime_harness.py | 9 | CODE | |
| LOW | tests/test_guidance_ttt_tasks.py | 3 | CODE | |
| LOW | tests/test_ceobench_harness.py | 3 | CODE | |
| LOW | tests/test_tttd_run_controller.py | 1 | CODE | |
| LOW | tests/test_coral_demo_grader.py | 9 | CODE | |
| LOW | tests/test_tttd_packing_tasks.py | 1 | CODE | |
| LOW | tests/test_coral_adapter.py | 9 | CODE | |
| LOW | tests/test_openclawrl_personas.py | 3 | CODE | |
| LOW | tests/test_reef_client.py | 3 | CODE | |
| LOW | tests/smoke/test_real_pi.py | 9 | CODE | |
| LOW | tests/smoke/test_real_terminus.py | 15 | CODE | |
| LOW | tests/smoke/test_real_codex.py | 9 | CODE | |
| LOW | tests/smoke/test_real_hermes.py | 11 | CODE | |
| LOW | tests/smoke/test_reefine_scripted_trial.py | 3 | CODE | |
| LOW | tests/smoke/test_real_dsh.py | 10 | CODE | |
| LOW | tests/reef_service/test_score_centering_parity.py | 15 | CODE | |
| LOW | tests/reef_service/test_harness_evolve_processor.py | 9 | CODE | |
| LOW | tests/reef_service/test_multi_scenario_training.py | 3 | CODE | |
| LOW | tests/reef_service/test_training_coordinator.py | 3 | CODE | |
| LOW | tests/reef_service/test_sglang_chat_capture.py | 1 | CODE | |
| LOW | tests/reef_service/test_startup_diagnostics.py | 3 | CODE | |
| LOW | tests/reef_service/test_serve_profile.py | 3 | CODE | |
| LOW | tests/reef_service/test_tinker_local_engine.py | 3 | CODE | |
| LOW | tests/reef_service/test_component_trainers.py | 3 | CODE | |
| LOW | tests/reef_service/test_distill_algorithm.py | 9 | CODE | |
| LOW | tests/reef_service/test_surfaces.py | 1 | CODE | |
| LOW | tests/reef_service/test_evolution_executor_ray.py | 3 | CODE | |
| LOW | tests/reef_service/runtime_stubs.py | 3 | CODE | |
| LOW | tests/reef_service/test_verifier_reward.py | 3 | CODE | |
| LOW | tests/reef_service/test_task_generation_processor.py | 3 | CODE | |
| LOW | tests/reef_service/test_harness_model_route.py | 11 | CODE | |
| LOW | tests/reef_service/test_record_store_contract.py | 3 | CODE | |
| LOW | tests/reef_service/test_failure_manifest.py | 7 | CODE | |
| LOW | tests/reef_service/test_scenario_release_concurrency.py | 3 | CODE | |
| LOW | tests/reef_service/test_distill_parity.py | 16 | CODE | |
| LOW | tests/reef_service/test_record2dataset_service.py | 3 | CODE | |
| LOW | tests/reef_service/test_runtime_registry.py | 1 | CODE | |
| LOW | tests/reef_service/test_terminus_tree.py | 8 | CODE | |
| LOW | tests/reef_service/test_provider_startup.py | 3 | CODE | |
| LOW | tests/reef_service/test_requests_extension.py | 21 | CODE | |
| LOW | tests/reef_service/test_wandb_tracking.py | 1 | CODE | |
| LOW | tests/reef_service/test_gepa.py | 6 | CODE | |
| LOW | tests/reef_service/test_release_components.py | 3 | CODE | |
| 1017 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | tests/test_openclawrl.py | 589 | worker.push(TurnJudgment("i0", score=1.0, teacher_cands=(cand,))) | CODE |
| HIGH | tests/test_openclawrl.py | 591 | worker.push(TurnJudgment("i1", score=1.0, teacher_cands=({"hint": "", "teacher_tokens": [1, 2, 3, 4]},))) | CODE |
| HIGH | tests/test_spade_training.py | 257 | "Dockerfile": "FROM python:3.12-slim\nRUN apt-get update && apt-get install -y tmux && echo 8471 > /var/run/app. | CODE |
| HIGH | tests/test_openclawrl_personas.py | 127 | worker.push( | CODE |
| HIGH⚡ | tests/smoke/test_reefine_scripted_trial.py | 32 | pi.on('before_agent_start', event => active ? {systemPrompt: 'Chat only.'} : undefined); | CODE |
| HIGH⚡ | tests/smoke/test_reefine_scripted_trial.py | 33 | pi.on('tool_call', event => active && event.toolName !== 'web_search' | CODE |
| HIGH⚡ | tests/smoke/test_reefine_scripted_trial.py | 34 | ? {block: true, reason: 'chat mode'} : undefined); | CODE |
| HIGH | tests/smoke/test_reefine_scripted_trial.py | 213 | ? {systemPrompt: event.systemPrompt + '\\nAnswer concisely.'} : undefined); | CODE |
| HIGH | tests/reef_service/test_record2dataset_service.py | 67 | "Dockerfile": "FROM python:3.12-slim\nRUN apt-get update && apt-get install -y tmux && echo 8471 > /var/run/app. | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 98 | sendUserMessage(text, options) { events.push({ kind: "user_message", text, options: options ?? null }); }, | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 98 | sendUserMessage(text, options) { events.push({ kind: "user_message", text, options: options ?? null }); }, | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 99 | sendMessage(message, options) { events.push({ kind: "message", message, options: options ?? null }); }, | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 99 | sendMessage(message, options) { events.push({ kind: "message", message, options: options ?? null }); }, | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 100 | appendEntry(customType, data) { events.push({ kind: "entry", customType, data }); }, | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 101 | registerEntryRenderer(customType) { entryRenderers.push(customType); }, | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 102 | exec: async (command, args) => { events.push({ kind: "exec", command, args }); return execAnswerFor(args); }, | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 115 | events.push({ kind: "model_call", model, context: JSON.parse(JSON.stringify(context)), signal: options.signal inst | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 117 | if (reply && reply.hang) return new Promise(() => {}); | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 118 | if (!reply || reply.throw) throw new Error(reply ? reply.throw : "no scripted reply"); | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 124 | confirm: async (title, message) => { events.push({ kind: "confirm", title, message }); return confirmAnswer(); }, | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 125 | notify: (message, type) => events.push({ kind: "notify", message, type }), | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 126 | select: async (title, options) => { events.push({ kind: "select", title, options }); return selections.shift() ?? un | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 126 | select: async (title, options) => { events.push({ kind: "select", title, options }); return selections.shift() ?? un | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 127 | input: async (title, placeholder) => { events.push({ kind: "input", title, placeholder }); return inputs.shift() ?? | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 127 | input: async (title, placeholder) => { events.push({ kind: "input", title, placeholder }); return inputs.shift() ?? | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 128 | setStatus: (key, text) => events.push({ kind: "status", key, text: text ?? null }), | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 128 | setStatus: (key, text) => events.push({ kind: "status", key, text: text ?? null }), | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 129 | setWidget: (key, content) => events.push({ kind: "widget", key, content: content ?? null }), | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 129 | setWidget: (key, content) => events.push({ kind: "widget", key, content: content ?? null }), | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 156 | const skewMs = Number(process.env.TEST_CLOCK_SKEW_MS || 31 * 60 * 1000); | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 161 | const out = { tools: Object.keys(tools), commands: Object.keys(commands), shortcuts: Object.keys(shortcuts), handlers: O | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 166 | const later = index > 0 && process.env.TEST_ARGS_2 !== undefined; | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 166 | const later = index > 0 && process.env.TEST_ARGS_2 !== undefined; | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 170 | if (step === "ask_user") return tools.reef_ask_user.execute("call-1", params, undefined, undefined, ctx); | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 171 | if (step === "file_request") return tools.reef_file_request.execute("call-1", params, undefined, undefined, ctx); | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 173 | return null; | CODE |
| HIGH⚡ | tests/reef_service/test_requests_extension.py | 177 | const repeats = Number(process.env.TEST_REPEAT || 1); | CODE |
| HIGH | tests/reef_service/test_requests_extension.py | 87 | const keys = Object.keys(execAnswers).filter((key) => joined === key || joined.startsWith(`${key} `)); | CODE |
| HIGH | tests/reef_service/test_requests_extension.py | 143 | events.push({ kind: "fetch", method, url, headers: init.headers ?? {}, body: init.body ? JSON.parse(init.body) : null, | CODE |
| HIGH | tests/reef_service/test_requests_extension.py | 143 | events.push({ kind: "fetch", method, url, headers: init.headers ?? {}, body: init.body ? JSON.parse(init.body) : null, | CODE |
| HIGH | tests/reef_service/test_requests_extension.py | 191 | events.push({ kind: "shortcut", key: Object.keys(shortcuts)[0] }); | CODE |
| HIGH | tests/reef_service/test_requests_extension.py | 195 | await sleep(Number(process.env.TEST_WAIT_MS || 0)); | CODE |
| HIGH | tests/reef_service/test_harness_example.py | 779 | " return process.env.TWILIO_AUTH_TOKEN && process.env.SMS_TO;\n" | CODE |
| HIGH | tests/reef_service/test_harbor_tasks.py | 23 | '[ "$(cat /workspace/answer.txt)" = 391 ] && echo 1 > /logs/verifier/reward.txt || echo 0 > /logs/verifier/reward.tx | CODE |
| HIGH | tests/reef_service/test_record2dataset_designer.py | 33 | "Dockerfile": "FROM python:3.12-slim\nRUN apt-get update && apt-get install -y tmux && echo 8471 > /var/run/app. | CODE |
| HIGH | tests/reef_service/test_records.py | 258 | artifact_json = '{"kind":"artifact","content_id":"内容","release_id":"发布","parent_release_id":null}' | CODE |
| HIGH⚡ | tests/reef_service/test_openclawrl.py | 185 | worker.push(TurnJudgment("declined")) # judge declined: no RL, no directive | CODE |
| HIGH⚡ | tests/reef_service/test_openclawrl.py | 192 | worker.push(TurnJudgment("ragged", score=-1.0, teacher_cands=ANCHOR)) | CODE |
| HIGH⚡ | tests/reef_service/test_openclawrl.py | 205 | worker.push(TurnJudgment("old", score=1.0, teacher_cands=ANCHOR)) | CODE |
| HIGH⚡ | tests/reef_service/test_openclawrl.py | 206 | worker.push(TurnJudgment("new", score=1.0, teacher_cands=ANCHOR)) | CODE |
| HIGH⚡ | tests/reef_service/test_openclawrl.py | 358 | worker.push(TurnJudgment("new", score=1.0, teacher_cands=ANCHOR)) | CODE |
| HIGH⚡ | tests/reef_service/test_openclawrl.py | 359 | worker.push(TurnJudgment("old", score=1.0, teacher_cands=ANCHOR)) | CODE |
| HIGH | tests/reef_service/test_openclawrl.py | 118 | worker.push(TurnJudgment("t1", score=1.0, teacher_cands=ANCHOR)) | CODE |
| HIGH | tests/reef_service/test_openclawrl.py | 479 | worker.push(TurnJudgment("t1", score=1.0, teacher_cands=ANCHOR)) | CODE |
| HIGH | tests/reef_service/test_record_tracing.py | 208 | 'data: {"choices":[{"delta":{"content":"hello"}}],"usage":null}\n\n' | CODE |
| HIGH⚡ | tests/reef_service/test_version_check.py | 84 | assert "checked || process.env.PI_OFFLINE" in text | CODE |
| HIGH⚡ | tests/reef_service/test_version_check.py | 206 | const keys = Object.keys(execAnswers).filter((key) => joined === key || joined.startsWith(`${key} `)); | STRING |
| HIGH⚡ | tests/reef_service/test_version_check.py | 215 | events.push({ kind: "exec", command, args }); | STRING |
| HIGH⚡ | tests/reef_service/test_version_check.py | 223 | events.push({ kind: "fetch", method: "POST", url, body: init.body ? JSON.parse(init.body) : null }); | STRING |
| HIGH⚡ | tests/reef_service/test_version_check.py | 223 | events.push({ kind: "fetch", method: "POST", url, body: init.body ? JSON.parse(init.body) : null }); | STRING |
| 88 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | .pre-commit-config.yaml | 11 | # are published as produced; harness golden trees are byte exact render | COMMENT |
| MEDIUM⚡ | tests/test_tttd_harness.py | 257 | # Reef's weight surface names the served adapter; the harness never does. | COMMENT |
| MEDIUM | tests/test_guidance_ttt.py | 400 | # As a Scorer, the same verdict becomes the harness's VerificationResult. | COMMENT |
| MEDIUM | tests/test_gepa_aime_harness.py | 31 | # Every example names its package ``harness`` and its driver ``run``: | COMMENT |
| MEDIUM⚡ | tests/test_ceobench_harness.py | 349 | # The benchmark's agent loop (harness/agent.py) | COMMENT |
| MEDIUM⚡ | tests/test_ceobench_harness.py | 675 | # The task container (harness/harbor_agent.py) | COMMENT |
| MEDIUM⚡ | tests/test_ceobench_harness.py | 755 | # Weeks, decisions, and the reward (harness/report.py) | COMMENT |
| MEDIUM⚡ | tests/test_ceobench_harness.py | 896 | # One episode through the Harbor agent (harness/harbor_agent.py) | COMMENT |
| MEDIUM | tests/test_ceobench_harness.py | 514 | # The tools (harness/tools.py): the executor script, run here the way the container runs it | COMMENT |
| MEDIUM | tests/test_ceobench_harness.py | 1128 | # The patch touches the engine only: the agent, its tools, and its runner live in harness/. | COMMENT |
| MEDIUM⚡ | tests/test_tttd_packing_tasks.py | 80 | # shared harness; instruction.md contains only the immutable task contract. | COMMENT |
| MEDIUM | tests/smoke/test_real_pi.py | 45 | pass # keep the harness's stderr out of pytest output | CODE |
| MEDIUM | tests/smoke/test_real_hermes.py | 54 | pass # keep the harness's stderr out of pytest output | CODE |
| MEDIUM | tests/smoke/test_real_dsh.py | 46 | pass # keep the harness's stderr out of pytest output | CODE |
| MEDIUM | tests/reef_service/test_serve_profile.py | 31 | # The folded harness-evolve profile keeps its name as an alias, not as a profile of its own. | COMMENT |
| MEDIUM | tests/reef_service/test_component_trainers.py | 442 | # The weights result was prepared against the base, which the harness commit replaced. | COMMENT |
| MEDIUM | tests/reef_service/test_component_trainers.py | 495 | # The harness step is mid evaluation: the dispatched trainer reserves, executes and commits meanwhile... | COMMENT |
| MEDIUM | tests/reef_service/test_component_trainers.py | 542 | # The job has not started: the harness cycle still holds the engine it evaluates through. | COMMENT |
| MEDIUM | tests/reef_service/test_component_trainers.py | 688 | # The next start finds both batches unread and trains them, with no new record to wake the harness. | COMMENT |
| MEDIUM | tests/reef_service/test_component_trainers.py | 1171 | # Its source trained, so the rebuilt harness skips it; the weights trainer still reads it. | COMMENT |
| MEDIUM | tests/reef_service/test_component_trainers.py | 1188 | # Settled, not live: commit B names it as consumed, so the rebuilt harness skips it. | COMMENT |
| MEDIUM | tests/reef_service/test_component_trainers.py | 1632 | # Only the harness changed: the weights job's commit still stands. | COMMENT |
| MEDIUM | tests/reef_service/test_component_trainers.py | 1748 | # The harness worker commits while the weights job is out. | COMMENT |
| MEDIUM | tests/reef_service/test_component_trainers.py | 1872 | # A rejected harness step is listed for its page, but it published nothing and names no head. | COMMENT |
| MEDIUM | tests/reef_service/test_component_trainers.py | 1891 | # The next harness release descends from h1 in the listed chain; the weights release it was published | COMMENT |
| MEDIUM | tests/reef_service/test_component_trainers.py | 2032 | # Back to the base: the seed tree is served again, and that is a harness release. | COMMENT |
| MEDIUM | tests/reef_service/test_component_trainers.py | 2086 | # The record store is gone; the checkpoint is adopted from the head and belongs to the harness trainer. | COMMENT |
| MEDIUM | tests/reef_service/test_component_trainers.py | 2161 | # The weights result is prepared against the served release; the harness cycle then replaces it. | COMMENT |
| MEDIUM | tests/reef_service/test_gepa.py | 40 | # The fake harness scores itself: its trajectory carries the rules text, so | COMMENT |
| MEDIUM⚡ | tests/reef_service/test_release_components.py | 390 | # The record names the harness trainer, but the tree did not change: no new harness head. | COMMENT |
| MEDIUM⚡ | tests/reef_service/test_release_components.py | 400 | # and live weights cannot carry the harness. | COMMENT |
| MEDIUM⚡ | tests/reef_service/test_release_components.py | 410 | # Rolling back to the harness-only step restores the base weights and evolved harness together. | COMMENT |
| MEDIUM⚡ | tests/reef_service/test_release_components.py | 418 | # Rolling back to the base keeps the served weights and only restores the seed harness. | COMMENT |
| MEDIUM⚡ | tests/reef_service/test_release_components.py | 424 | # A harness step held for review is promoted onto the weights served by then, not the weights of its day. | COMMENT |
| MEDIUM | tests/reef_service/test_release_components.py | 290 | # One trainer, bound to the harness, publishes whichever component a step names. | COMMENT |
| MEDIUM | tests/reef_service/test_release_components.py | 367 | # A harness step carries the weights forward and does not reload them. | COMMENT |
| MEDIUM | tests/reef_service/test_release_components.py | 378 | # A weights step activates the new weights and keeps the evolved harness. | COMMENT |
| MEDIUM | tests/reef_service/test_release_components.py | 443 | # Weights held for review by the harness trainer are promoted as weights, onto the tree served by then. | COMMENT |
| MEDIUM | tests/reef_service/test_release_components.py | 458 | # That promote changed no tree: the harness head and catalog stay at the newest harness step. | COMMENT |
| MEDIUM | tests/reef_service/test_release_components.py | 535 | # Newest first: the rejected step's row (named by the first harness release), that release, the creation. | COMMENT |
| MEDIUM | tests/reef_service/test_auth.py | 29 | # The two harness pages a browser opens by a link, and their neighbours that are no page. | COMMENT |
| MEDIUM | tests/reef_service/test_skillclaw_catalog.py | 59 | # Every example names its package ``harness``: drop any sibling | STRING |
| MEDIUM | tests/reef_service/test_composite_recipe.py | 499 | # The config component runs no step: the one commit is the harness's, naming the rows it consumed. | COMMENT |
| MEDIUM | tests/reef_service/test_composite_recipe.py | 609 | # The component with no report contract is listed first; the harness contract still guards ingress. | COMMENT |
| MEDIUM | tests/reef_service/test_harness_example.py | 55 | # Every example names its package ``harness``: drop any sibling | COMMENT |
| MEDIUM⚡ | tests/reef_service/test_skillclaw.py | 378 | # The pool reaches traffic through the proxy; no harness tree is installed, so no adapter is named. | COMMENT |
| MEDIUM | tests/reef_service/test_skillclaw.py | 110 | # Every example names its package ``harness``: drop any sibling | STRING |
| MEDIUM | tests/reef_service/test_composite_restart_replay.py | 23 | MOVED_TO = 3 # three harness commits landed while the job was out | CODE |
| MEDIUM | tests/reef_service/test_composite_restart_replay.py | 94 | # After the restart the trainer rebuilds the same batch; the scenario step is what the harness left. | COMMENT |
| MEDIUM⚡ | tests/reef_service/test_meta_harness.py | 272 | # Meta-Harness improves a fixed model's harness with a separate proposer. | COMMENT |
| MEDIUM⚡ | tests/reef_service/test_meta_harness.py | 274 | # turn the harness under test into its own author, measuring a different | COMMENT |
| MEDIUM | tests/reef_service/test_harness_step_record.py | 38 | # A harness shaped like the native loop: the root session, one delegated checker | COMMENT |
| MEDIUM | tests/reef_service/test_harness_recipe.py | 49 | # The fake harness scores itself: its trajectory carries the rules text, so | COMMENT |
| MEDIUM | tests/reef_service/test_harness_channel.py | 61 | # The fake harness scores itself, as in test_harness_recipe.py: its | COMMENT |
| MEDIUM | tests/reef_service/test_harness_channel.py | 650 | # --- the one-command install script (GET /reef/harness/install) --- | COMMENT |
| MEDIUM | tests/reef_service/test_harness_channel.py | 821 | #: The bundled adapters whose descriptor declares an install section: the ones GET /reef/harness/install serves. | COMMENT |
| MEDIUM | tests/reef_service/test_harness_channel.py | 929 | # Reef harness install: adapter pi, release v1. | COMMENT |
| MEDIUM | tests/reef_service/test_harness_channel.py | 930 | # Self contained: the composition files ride inline below and the harness | COMMENT |
| MEDIUM | tests/reef_service/test_harness_channel.py | 1216 | # Generated by reef harness install (adapter pi, release v1). | COMMENT |
| MEDIUM | tests/reef_service/test_harness_channel.py | 1219 | # reef-pi evolve "what the harness should do" # ask reef for a change | COMMENT |
| 160 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | recipes/beta/coral/reporter.py | 19 | logger = logging.getLogger(__name__) | CODE |
| LOW | recipes/beta/coral/__init__.py | 23 | __all__ = [ | CODE |
| LOW | recipes/beta/coral/processor.py | 32 | logger = logging.getLogger(__name__) | CODE |
| LOW | recipes/beta/coral/watcher.py | 51 | logger = logging.getLogger(__name__) | CODE |
| LOW | recipes/beta/coral/middleware.py | 22 | logger = logging.getLogger(__name__) | CODE |
| LOW | recipes/beta/spade/__init__.py | 31 | __all__ = [ | CODE |
| LOW | recipes/beta/spade/processor.py | 66 | logger = logging.getLogger(__name__) | CODE |
| LOW | recipes/sdpo/__init__.py | 26 | __all__ = ["SDPOProcessor", "SDPORecipe", "SDPOReport", "SdpoObjective"] | CODE |
| LOW | recipes/sdpo/processor.py | 193 | __all__ = ["FEEDBACK_TEMPLATE", "REPROMPT_TEMPLATE", "SOLUTION_TEMPLATE", "SDPOProcessor"] | CODE |
| LOW | recipes/sdpo/report.py | 21 | __all__ = ["SDPOReport"] | CODE |
| LOW | recipes/sdpo/slime/__init__.py | 65 | __all__ = ["SdpoAlgorithm", "SdpoSettings"] | CODE |
| LOW | recipes/sdpo/examples/sciknoweval/harness/__init__.py | 6 | __all__ = ["HarborAgent"] | CODE |
| LOW | recipes/gepa/backend.py | 82 | __all__ = ["ARCHIVE_STATE_KEY", "GEPABackend"] | CODE |
| LOW | recipes/sdft/__init__.py | 24 | __all__ = ["SDFTProcessor", "SDFTRecipe", "SdftObjective"] | CODE |
| LOW | recipes/sdft/processor.py | 58 | __all__ = ["CONTEXT_PLACEHOLDER", "DEFAULT_CONTEXT_TEMPLATE", "SDFTProcessor", "context_block_template"] | CODE |
| LOW | recipes/sdft/slime/__init__.py | 52 | __all__ = ["SdftAlgorithm", "SdftSettings"] | CODE |
| LOW | recipes/sdft/examples/skill_stream/harness/__init__.py | 6 | __all__ = ["HarborAgent"] | CODE |
| LOW | recipes/openclawrl/prm.py | 30 | logger = logging.getLogger(__name__) | CODE |
| LOW | recipes/openclawrl/__init__.py | 23 | __all__ = ["OpenClawRLObjective", "OpenClawRLProcessor", "OpenClawRLRecipe"] | CODE |
| LOW | recipes/openclawrl/processor.py | 52 | logger = logging.getLogger(__name__) | CODE |
| LOW | recipes/openclawrl/slime/objective.py | 34 | logger = logging.getLogger(__name__) | CODE |
| LOW | recipes/openclawrl/slime/teacher.py | 50 | logger = logging.getLogger(__name__) | CODE |
| LOW | recipes/openclawrl/slime/teacher.py | 208 | __all__ = [ | CODE |
| LOW | recipes/openclawrl/slime/__init__.py | 159 | __all__ = ["OpenclawrlAlgorithm", "OpenclawrlSettings"] | CODE |
| LOW | …pes/openclawrl/examples/openclawrl/harness/__init__.py | 10 | __all__ = ["HermesStreamAgent"] | CODE |
| LOW | recipes/basic/harness/__init__.py | 6 | __all__ = ["HarborAgent"] | CODE |
| LOW | recipes/skillclaw/run.py | 467 | def _settle(service: RunService) -> None: | CODE |
| LOW | recipes/skillclaw/harness/day.py | 206 | def _setup_workspace(container: str) -> None: | CODE |
| LOW | recipes/tttd/__init__.py | 28 | __all__ = ["TTTDGroupedRolloutReport", "TTTDProcessor", "TTTDRecipe", "TttdObjective"] | CODE |
| LOW | recipes/tttd/processor.py | 15 | logger = logging.getLogger(__name__) | CODE |
| LOW | recipes/tttd/report.py | 7 | __all__ = ["TTTDGroupedRolloutReport"] | CODE |
| LOW | recipes/tttd/slime/__init__.py | 37 | __all__ = ["TttdAlgorithm"] | CODE |
| LOW | recipes/tttd/examples/guidance_ttt/judges/protocol.py | 113 | __all__ = ["parse_submission", "serve"] | CODE |
| LOW | recipes/tttd/examples/guidance_ttt/harness/library.py | 705 | def _update_puct_stats_for_group(self, group: dict[str, Any]) -> None: | CODE |
| LOW | recipes/tttd/examples/guidance_ttt/harness/__init__.py | 27 | __all__ = [ | CODE |
| LOW | …s/tttd/examples/guidance_ttt/harness/run_controller.py | 443 | __all__ = [ | CODE |
| LOW | recipes/tttd/examples/guidance_ttt/harness/search.py | 51 | __all__ = ["guidance_chat_request", "openai_action"] | CODE |
| LOW | recipes/tttd/examples/guidance_ttt/harness/contract.py | 64 | __all__ = ["SCORE_DIRECTIONS", "TaskContract"] | CODE |
| LOW | recipes/tttd/examples/guidance_ttt/harness/scorer.py | 255 | __all__ = [ | CODE |
| LOW | recipes/tttd/examples/tttd/harness/__init__.py | 17 | __all__ = ["HarborAgent", "TTTDChatRequestBuilder", "openai_action"] | CODE |
| LOW | recipes/tttd/examples/tttd/harness/run_controller.py | 438 | __all__ = [ | CODE |
| LOW | recipes/meta_harness/backend.py | 142 | __all__ = ["POPULATION_STATE_KEY", "MetaHarnessBackend"] | CODE |
| LOW | recipes/meta_harness/population.py | 390 | __all__ = [ | CODE |
| LOW | recipes/meta_harness/method.py | 441 | __all__ = ["SEARCH_MODES", "MetaHarnessPlugin", "MetaHarnessProposer", "MetaHarnessSelectorMixin", "mutations_between"] | CODE |
| LOW | recipes/meta_harness/recipe.py | 226 | __all__ = ["MetaHarnessRecipe", "scenario_population_path"] | CODE |
| LOW | recipes/sao/__init__.py | 18 | __all__ = ["SAOProcessor", "SAORecipe", "SaoObjective"] | CODE |
| LOW | recipes/sao/slime/__init__.py | 299 | __all__ = ["SaoAlgorithm", "SaoSettings"] | CODE |
| LOW | recipes/sao/examples/ceobench/harness/__init__.py | 10 | __all__ = ["HarborAgent"] | CODE |
| LOW | …cipes/sao/examples/imo_answerbench/harness/__init__.py | 8 | __all__ = ["HarborAgent"] | CODE |
| LOW | reef/dispatcher.py | 48 | logger = logging.getLogger(__name__) | CODE |
| LOW | reef/dispatcher.py | 1333 | def _set_training_storage_status(self, value: Mapping[str, Any] | None) -> None: | CODE |
| LOW | reef/__init__.py | 48 | __all__ = [ | CODE |
| LOW | reef/core/artifact_ref.py | 125 | __all__ = [ | CODE |
| LOW | reef/core/evaluation.py | 108 | __all__ = [ | CODE |
| LOW | reef/core/__init__.py | 31 | __all__ = [ | CODE |
| LOW | reef/core/requirements.py | 156 | __all__ = [ | CODE |
| LOW | reef/core/components.py | 138 | __all__ = [ | CODE |
| LOW | reef/core/tasks/__init__.py | 21 | __all__ = [ | CODE |
| LOW | reef/core/reports/__init__.py | 17 | __all__ = [ | CODE |
| LOW | reef/core/reports/scored_rollout.py | 7 | __all__ = ["ScoredRolloutReport"] | CODE |
| 172 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tests/test_spade_training.py | 447 | except Exception: | CODE |
| LOW | tests/test_spade_training.py | 487 | except Exception: | CODE |
| MEDIUM | tests/reef_service/test_component_trainers.py | 641 | def cycle() -> None: | CODE |
| LOW | tests/reef_service/test_component_trainers.py | 644 | except Exception as exc: | CODE |
| MEDIUM | tests/reef_service/test_release_components.py | 607 | def read() -> None: | CODE |
| LOW | tests/reef_service/test_release_components.py | 610 | except Exception as exc: | CODE |
| LOW | tests/reef_service/test_executor.py | 110 | except Exception as exc: | CODE |
| LOW | tests/reef_service/test_reef_concurrency.py | 105 | except Exception as exc: | CODE |
| MEDIUM | tests/reef_service/test_commit_log.py | 1072 | def accept_report() -> None: | CODE |
| MEDIUM | tests/reef_service/test_commit_log.py | 1088 | def read_status() -> None: | CODE |
| LOW | tests/reef_service/test_commit_log.py | 1077 | except Exception as exc: | CODE |
| LOW | tests/reef_service/test_commit_log.py | 1091 | except Exception as exc: | CODE |
| LOW | tests/reef_service/test_commit_log.py | 1388 | except Exception as exc: | CODE |
| LOW | recipes/beta/coral/examples/coral_demo/run.py | 54 | except Exception: | CODE |
| LOW | …beta/coral/examples/coral_demo/smoke/scripted_agent.py | 47 | except Exception as error: # gateway warming up / transient — retry, never crash-loop | CODE |
| LOW | …coral_demo/task/grader/src/coral_demo_grader/grader.py | 85 | except Exception: | CODE |
| LOW | …coral_demo/task/grader/src/coral_demo_grader/grader.py | 91 | except Exception: | CODE |
| LOW | …coral_demo/task/grader/src/coral_demo_grader/grader.py | 110 | except Exception: | CODE |
| LOW | …coral_demo/task/grader/src/coral_demo_grader/grader.py | 118 | except Exception: | CODE |
| MEDIUM | …coral_demo/task/grader/src/coral_demo_grader/grader.py | 88 | def check(a, b, expected) -> bool: | CODE |
| LOW | …kill_stream/harbor/tooluse/environment/judge_server.py | 134 | except Exception as e: # a broken submission scores 0, loudly | CODE |
| LOW | …kill_stream/harbor/science/environment/judge_server.py | 134 | except Exception as e: # a broken submission scores 0, loudly | CODE |
| LOW | recipes/openclawrl/prm.py | 339 | except Exception: | CODE |
| LOW | recipes/openclawrl/prm.py | 391 | except Exception: | CODE |
| LOW | recipes/openclawrl/prm.py | 433 | except Exception: | CODE |
| LOW | …nclawrl/examples/openclawrl/user_sim/student_server.py | 129 | except Exception: # a judging outage must not wedge the session | CODE |
| LOW | …nclawrl/examples/openclawrl/user_sim/student_server.py | 201 | except Exception: | CODE |
| LOW | …nclawrl/examples/openclawrl/user_sim/student_server.py | 249 | except Exception: # a raised handler closes the socket with zero | CODE |
| LOW | …enclawrl/examples/openclawrl/results/learning_curve.py | 238 | except Exception: | CODE |
| LOW | recipes/skillclaw/run.py | 392 | except Exception: | CODE |
| LOW | recipes/skillclaw/run.py | 461 | except Exception: | CODE |
| LOW | recipes/skillclaw/serving/judge_proxy.py | 87 | except Exception as err: | CODE |
| LOW | recipes/skillclaw/serving/judge_proxy.py | 114 | except Exception: | CODE |
| LOW⚡ | recipes/skillclaw/harbor/tests/grade.py | 49 | except Exception as e: | CODE |
| LOW | recipes/skillclaw/harbor/tests/grade.py | 227 | except Exception: | CODE |
| LOW | recipes/skillclaw/harbor/tests/grade.py | 321 | except Exception: | CODE |
| LOW | recipes/skillclaw/harbor/tests/grade.py | 351 | except Exception: | CODE |
| MEDIUM | recipes/skillclaw/harbor/tests/grade.py | 44 | def fetch_audit(url): | CODE |
| LOW | recipes/skillclaw/harness/night.py | 56 | except Exception: | CODE |
| LOW | recipes/skillclaw/harness/night.py | 244 | except Exception as exc: | CODE |
| LOW | recipes/skillclaw/harness/night.py | 256 | except Exception as exc: | CODE |
| LOW | recipes/skillclaw/harness/sessions.py | 50 | except Exception: | CODE |
| LOW | recipes/skillclaw/harness/evolver.py | 86 | except Exception as exc: | CODE |
| LOW | recipes/skillclaw/harness/evolver.py | 284 | except Exception: | CODE |
| LOW | recipes/skillclaw/harness/evolver.py | 383 | except Exception: | CODE |
| LOW | recipes/skillclaw/harness/day.py | 576 | except Exception as exc: | CODE |
| LOW | recipes/skillclaw/harness/day.py | 659 | except Exception as exc: | CODE |
| LOW | …s/guidance_ttt/harbor/polyomino_packing/tests/grade.py | 89 | except Exception as error: # An unreachable judge is a zero, not a crash. | CODE |
| LOW | …ttd/examples/guidance_ttt/harbor/ahc058/tests/grade.py | 89 | except Exception as error: # An unreachable judge is a zero, not a crash. | CODE |
| LOW | …ttd/examples/guidance_ttt/harbor/trimul/tests/grade.py | 89 | except Exception as error: # An unreachable judge is a zero, not a crash. | CODE |
| LOW | …examples/guidance_ttt/harbor/lasso_path/tests/grade.py | 89 | except Exception as error: # An unreachable judge is a zero, not a crash. | CODE |
| LOW | …/tttd/examples/guidance_ttt/judges/ale_bench_server.py | 121 | except Exception: | CODE |
| LOW | recipes/tttd/examples/guidance_ttt/judges/protocol.py | 58 | except Exception as exc: | CODE |
| LOW | recipes/tttd/examples/guidance_ttt/judges/protocol.py | 97 | except Exception as exc: | CODE |
| LOW | recipes/tttd/examples/guidance_ttt/harness/agent.py | 275 | except Exception as exc: # External execution is an environment outcome. | CODE |
| LOW | recipes/tttd/examples/guidance_ttt/harness/state.py | 16 | except Exception: | CODE |
| LOW | …/examples/tttd/harbor/circle_packing_32/tests/grade.py | 27 | except Exception as e: | CODE |
| LOW | …d/harbor/circle_packing_32/environment/judge_server.py | 119 | except Exception as e: # a broken submission scores 0, loudly | CODE |
| LOW | …d/harbor/circle_packing_32/environment/judge_server.py | 169 | except Exception as e: | CODE |
| LOW | …les/tttd/harbor/circle_packing_32/environment/score.py | 140 | except Exception as exc: | CODE |
| 147 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | tests/reef_service/test_gepa.py | 0 | \ #!/usr/bin/env python3 import json, os, sys from pathlib import path agent_dir = path(os.environ["pi_coding_agent_dir" | STRING |
| HIGH | tests/reef_service/test_skillclaw.py | 0 | \ #!/usr/bin/env python3 import json, os, sys from pathlib import path agent_dir = path(os.environ["pi_coding_agent_dir" | STRING |
| HIGH | tests/reef_service/test_harness_recipe.py | 0 | \ #!/usr/bin/env python3 import json, os, sys from pathlib import path agent_dir = path(os.environ["pi_coding_agent_dir" | STRING |
| HIGH | tests/reef_service/test_harness_channel.py | 0 | \ #!/usr/bin/env python3 import json, os, sys from pathlib import path agent_dir = path(os.environ["pi_coding_agent_dir" | STRING |
| HIGH | tests/reef_service/test_meta_harness.py | 0 | \ #!/usr/bin/env python3 import json, os, sys from pathlib import path agent_dir = path(os.environ["pi_coding_agent_dir" | STRING |
| HIGH | tests/reef_service/test_gepa.py | 0 | stands in for the training backend: these cases exercise ``decide`` only. | STRING |
| HIGH | tests/reef_service/test_meta_harness.py | 0 | stands in for the training backend: these cases exercise ``decide`` only. | STRING |
| HIGH | tests/reef_service/test_selection.py | 0 | stands in for the training backend: these cases exercise ``decide`` only. | STRING |
| HIGH | tests/reef_service/test_cordis_floor.py | 0 | stands in for the training backend: these cases exercise ``decide`` only. | STRING |
| HIGH | …/sciknoweval/harbor/chemistry/environment/chemistry.py | 0 | training releases committed so far; ``none`` while the service is busy, 0 before the scenario exists. | STRING |
| HIGH | …ples/skill_stream/harbor/tooluse/environment/skills.py | 0 | training releases committed so far; ``none`` while the service is busy, 0 before the scenario exists. | STRING |
| HIGH | …ples/skill_stream/harbor/science/environment/skills.py | 0 | training releases committed so far; ``none`` while the service is busy, 0 before the scenario exists. | STRING |
| HIGH | …/sciknoweval/harbor/chemistry/environment/chemistry.py | 0 | block until the scenario has committed ``expected`` training releases; return the count seen. | STRING |
| HIGH | …ples/skill_stream/harbor/tooluse/environment/skills.py | 0 | block until the scenario has committed ``expected`` training releases; return the count seen. | STRING |
| HIGH | …ples/skill_stream/harbor/science/environment/skills.py | 0 | block until the scenario has committed ``expected`` training releases; return the count seen. | STRING |
| HIGH | recipes/basic/harness/agent.py | 0 | nothing to install: the agent logic runs on the host. | STRING |
| HIGH | recipes/skillclaw/harness/harbor_agent.py | 0 | nothing to install: the agent logic runs on the host. | STRING |
| HIGH | recipes/sao/examples/imo_answerbench/harness/agent.py | 0 | nothing to install: the agent logic runs on the host. | STRING |
| HIGH | …s/guidance_ttt/harbor/polyomino_packing/tests/grade.py | 0 | the verifier: re-submit the trial's candidate to task judge and record it. the agent's own search already scored thousan | STRING |
| HIGH | …ttd/examples/guidance_ttt/harbor/ahc058/tests/grade.py | 0 | the verifier: re-submit the trial's candidate to task judge and record it. the agent's own search already scored thousan | STRING |
| HIGH | …ttd/examples/guidance_ttt/harbor/trimul/tests/grade.py | 0 | the verifier: re-submit the trial's candidate to task judge and record it. the agent's own search already scored thousan | STRING |
| HIGH | …examples/guidance_ttt/harbor/lasso_path/tests/grade.py | 0 | the verifier: re-submit the trial's candidate to task judge and record it. the agent's own search already scored thousan | STRING |
| HIGH | …/examples/tttd/harbor/circle_packing_32/instruction.md | 0 | validate that circles don't overlap and are inside the unit square args: centers: np.array of shape (n, 2) with (x, y) c | STRING |
| HIGH | …les/tttd/harbor/circle_packing_32/environment/score.py | 0 | validate that circles don't overlap and are inside the unit square args: centers: np.array of shape (n, 2) with (x, y) c | STRING |
| HIGH | …/examples/tttd/harbor/circle_packing_26/instruction.md | 0 | validate that circles don't overlap and are inside the unit square args: centers: np.array of shape (n, 2) with (x, y) c | STRING |
| HIGH | …les/tttd/harbor/circle_packing_26/environment/score.py | 0 | validate that circles don't overlap and are inside the unit square args: centers: np.array of shape (n, 2) with (x, y) c | STRING |
| HIGH | …ults/formal-8x64-v3-packing/packing26/best_solution.py | 0 | validate that circles don't overlap and are inside the unit square args: centers: np.array of shape (n, 2) with (x, y) c | STRING |
| HIGH | …/examples/tttd/harbor/circle_packing_32/tests/grade.py | 0 | the verifier's whole job under the judge protocol: ask the judge for the final result and write it down. generic — task | STRING |
| HIGH | …/examples/tttd/harbor/circle_packing_26/tests/grade.py | 0 | the verifier's whole job under the judge protocol: ask the judge for the final result and write it down. generic — task | STRING |
| HIGH | …/examples/tttd/harbor/erdos_min_overlap/tests/grade.py | 0 | the verifier's whole job under the judge protocol: ask the judge for the final result and write it down. generic — task | STRING |
| HIGH | …d/harbor/circle_packing_32/environment/judge_server.py | 0 | the judge: one http server, one scoring implementation, a submission budget. generic — task authors never edit this file | STRING |
| HIGH | …d/harbor/circle_packing_26/environment/judge_server.py | 0 | the judge: one http server, one scoring implementation, a submission budget. generic — task authors never edit this file | STRING |
| HIGH | …d/harbor/erdos_min_overlap/environment/judge_server.py | 0 | the judge: one http server, one scoring implementation, a submission budget. generic — task authors never edit this file | STRING |
| HIGH | …o/examples/imo_answerbench/harbor/imo-4/tests/grade.py | 0 | harbor verifier for imo problem_idx 8 (gold: $-\\frac{2023}{2024^2}$). extracts \\boxed{} from /workspace/answer.txt and | STRING |
| HIGH | …/examples/imo_answerbench/harbor/imo-12/tests/grade.py | 0 | harbor verifier for imo problem_idx 8 (gold: $-\\frac{2023}{2024^2}$). extracts \\boxed{} from /workspace/answer.txt and | STRING |
| HIGH | …o/examples/imo_answerbench/harbor/imo-8/tests/grade.py | 0 | harbor verifier for imo problem_idx 8 (gold: $-\\frac{2023}{2024^2}$). extracts \\boxed{} from /workspace/answer.txt and | STRING |
| HIGH | reef/runtime/interfaces.py | 0 | load a peft adapter directory the engine's host can read, under ``lora_name``. | STRING |
| HIGH | reef/inference/vllm/engine.py | 0 | load a peft adapter directory the engine's host can read, under ``lora_name``. | STRING |
| HIGH | reef/inference/sglang/engine.py | 0 | load a peft adapter directory the engine's host can read, under ``lora_name``. | STRING |
| HIGH | reef/harness/adapters/dsh/quirks.py | 0 | refuse a provider, an endpoint, a credential or a model the binding did not write. | STRING |
| HIGH | reef/harness/adapters/pi/quirks.py | 0 | refuse a provider, an endpoint, a credential or a model the binding did not write. | STRING |
| HIGH | reef/harness/adapters/hermes/quirks.py | 0 | refuse a provider, an endpoint, a credential or a model the binding did not write. | STRING |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tests/test_openclawrl_stream.py | 222 | CODE | |
| LOW | tests/smoke/test_real_codex.py | 394 | CODE | |
| LOW | tests/reef_service/test_scenario_release_concurrency.py | 208 | CODE | |
| LOW | tests/reef_service/test_record2dataset_service.py | 735 | CODE | |
| LOW | tests/reef_service/test_provider_startup.py | 233 | CODE | |
| LOW | tests/reef_service/test_slime_algorithm_contract.py | 425 | CODE | |
| LOW | tests/reef_service/test_slime_algorithm_contract.py | 481 | CODE | |
| LOW | tests/reef_service/test_slime_algorithm_contract.py | 645 | CODE | |
| LOW | tests/reef_service/test_dependency_boundaries.py | 26 | CODE | |
| LOW | tests/reef_service/test_dependency_boundaries.py | 357 | CODE | |
| LOW | tests/reef_service/test_dependency_boundaries.py | 394 | CODE | |
| LOW | tests/reef_service/test_dependency_boundaries.py | 41 | CODE | |
| LOW | tests/reef_service/test_service_executor_ray.py | 131 | CODE | |
| LOW | tests/reef_service/test_harness_channel.py | 95 | CODE | |
| LOW | tests/reef_service/test_task_player.py | 33 | CODE | |
| LOW | tests/reef_service/test_task_player.py | 39 | CODE | |
| LOW | tests/reef_service/test_native_harness.py | 1028 | CODE | |
| LOW | tests/reef_service/test_native_harness.py | 1038 | CODE | |
| LOW | tests/reef_service/test_postgres_records.py | 128 | CODE | |
| LOW | tests/reef_service/test_step_record_history.py | 83 | CODE | |
| LOW | tests/reef_service/test_step_record_history.py | 114 | CODE | |
| LOW | tests/reef_service/test_harness_wrapper.py | 2046 | CODE | |
| LOW | tests/reef_service/test_harness_wrapper.py | 2498 | CODE | |
| LOW | tests/reef_service/test_harness_wrapper.py | 2506 | CODE | |
| LOW | tests/reef_service/test_e2b_tunnel.py | 281 | CODE | |
| LOW | tests/reef_service/test_e2b_tunnel.py | 127 | CODE | |
| LOW | tests/reef_service/test_e2b_tunnel.py | 135 | CODE | |
| LOW | tests/reef_service/test_local_inference_startup.py | 150 | CODE | |
| LOW | tests/reef_service/test_training_backends.py | 141 | CODE | |
| LOW | recipes/beta/coral/middleware.py | 102 | CODE | |
| LOW | recipes/beta/coral/middleware.py | 141 | CODE | |
| LOW | recipes/gepa/archive.py | 275 | CODE | |
| LOW | recipes/gepa/examples/aime/run.py | 273 | CODE | |
| LOW | …kill_stream/harbor/tooluse/environment/judge_server.py | 203 | CODE | |
| LOW | …kill_stream/harbor/science/environment/judge_server.py | 203 | CODE | |
| LOW | recipes/openclawrl/slime/objective.py | 449 | CODE | |
| LOW | recipes/openclawrl/slime/teacher.py | 108 | CODE | |
| LOW | …enclawrl/examples/openclawrl/results/learning_curve.py | 161 | CODE | |
| LOW | recipes/skillclaw/run.py | 288 | CODE | |
| LOW | recipes/skillclaw/run.py | 553 | CODE | |
| LOW | recipes/skillclaw/serving/judge_proxy.py | 72 | CODE | |
| LOW | recipes/skillclaw/serving/judge_proxy.py | 132 | CODE | |
| LOW | recipes/skillclaw/harbor/tests/grade.py | 17 | CODE | |
| LOW | recipes/skillclaw/harness/sessions.py | 107 | CODE | |
| LOW | recipes/skillclaw/harness/day.py | 373 | CODE | |
| LOW | recipes/skillclaw/harness/skillclaw.py | 71 | CODE | |
| LOW | recipes/tttd/objective.py | 23 | CODE | |
| LOW | …pes/tttd/examples/guidance_ttt/judges/trimul_runner.py | 179 | CODE | |
| LOW | …/tttd/examples/guidance_ttt/judges/ale_bench_server.py | 93 | CODE | |
| LOW | recipes/tttd/examples/guidance_ttt/harness/library.py | 23 | CODE | |
| LOW | recipes/tttd/examples/guidance_ttt/harness/library.py | 287 | CODE | |
| LOW | recipes/tttd/examples/guidance_ttt/harness/library.py | 335 | CODE | |
| LOW | recipes/tttd/examples/guidance_ttt/harness/library.py | 399 | CODE | |
| LOW | recipes/tttd/examples/guidance_ttt/harness/scorer.py | 116 | CODE | |
| LOW | recipes/sao/examples/ceobench/harness/agent.py | 244 | CODE | |
| LOW | recipes/sao/examples/ceobench/harness/report.py | 284 | CODE | |
| LOW | recipes/sao/examples/imo_answerbench/evaluate.py | 110 | CODE | |
| LOW | .github/scripts/check_python_design.py | 96 | CODE | |
| LOW | .github/scripts/check_python_design.py | 242 | CODE | |
| LOW | .github/scripts/check_community_files.py | 134 | CODE | |
| 129 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | docker/patch/mcore_parse_hybrid_pattern.py | 2 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | docker/patch/mcore_parse_hybrid_pattern.py | 12 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_ceobench_harness.py | 348 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_ceobench_harness.py | 674 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_ceobench_harness.py | 754 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_ceobench_harness.py | 895 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/test_ceobench_harness.py | 513 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/test_ceobench_harness.py | 1042 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_coral_adapter.py | 87 | # -------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_coral_adapter.py | 89 | # -------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_coral_adapter.py | 136 | # -------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_coral_adapter.py | 138 | # -------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_coral_adapter.py | 161 | # -------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_coral_adapter.py | 163 | # -------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_coral_adapter.py | 232 | # -------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_coral_adapter.py | 234 | # -------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/reef_service/test_slime_algorithm_contract.py | 363 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/reef_service/test_slime_algorithm_contract.py | 366 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | recipes/openclawrl/slime/objective.py | 137 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | recipes/openclawrl/slime/objective.py | 139 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | recipes/openclawrl/slime/objective.py | 338 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | recipes/openclawrl/slime/objective.py | 340 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | recipes/openclawrl/slime/objective.py | 421 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | recipes/openclawrl/slime/objective.py | 423 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | …es/openclawrl/examples/openclawrl/user_sim/personas.py | 20 | # ─── student: wants non-AI-looking homework ────────────────────── | COMMENT |
| MEDIUM⚡ | recipes/skillclaw/harbor/tests/grade.py | 55 | # ══════════════════════════════════════════════════════════ | COMMENT |
| MEDIUM⚡ | recipes/skillclaw/harbor/tests/grade.py | 57 | # ══════════════════════════════════════════════════════════ | COMMENT |
| MEDIUM | recipes/skillclaw/harbor/tests/grade.py | 191 | # ══════════════════════════════════════════════════════════ | COMMENT |
| MEDIUM | recipes/skillclaw/harbor/tests/grade.py | 193 | # ══════════════════════════════════════════════════════════ | COMMENT |
| MEDIUM⚡ | …td/examples/guidance_ttt/solutions/trimul_qwen3_14b.py | 34 | # ════════════════════════════════════════════════════════════════════════ | COMMENT |
| MEDIUM⚡ | …td/examples/guidance_ttt/solutions/trimul_qwen3_14b.py | 36 | # ════════════════════════════════════════════════════════════════════════ | COMMENT |
| MEDIUM | …td/examples/guidance_ttt/solutions/trimul_qwen3_14b.py | 123 | # ════════════════════════════════════════════════════════════════════════ | COMMENT |
| MEDIUM | …td/examples/guidance_ttt/solutions/trimul_qwen3_14b.py | 126 | # ════════════════════════════════════════════════════════════════════════ | COMMENT |
| MEDIUM | …td/examples/guidance_ttt/solutions/trimul_qwen3_14b.py | 195 | # ════════════════════════════════════════════════════════════════════════ | COMMENT |
| MEDIUM | …td/examples/guidance_ttt/solutions/trimul_qwen3_14b.py | 197 | # ════════════════════════════════════════════════════════════════════════ | COMMENT |
| MEDIUM | …td/examples/guidance_ttt/solutions/trimul_qwen3_14b.py | 253 | # ════════════════════════════════════════════════════════════════════════ | COMMENT |
| MEDIUM | …td/examples/guidance_ttt/solutions/trimul_qwen3_14b.py | 256 | # ════════════════════════════════════════════════════════════════════════ | COMMENT |
| MEDIUM | …td/examples/guidance_ttt/solutions/trimul_qwen3_14b.py | 317 | # ════════════════════════════════════════════════════════════════════════ | COMMENT |
| MEDIUM | …td/examples/guidance_ttt/solutions/trimul_qwen3_14b.py | 319 | # ════════════════════════════════════════════════════════════════════════ | COMMENT |
| MEDIUM | …td/examples/guidance_ttt/solutions/trimul_qwen3_14b.py | 372 | # ════════════════════════════════════════════════════════════════════════ | COMMENT |
| MEDIUM | …td/examples/guidance_ttt/solutions/trimul_qwen3_14b.py | 374 | # ════════════════════════════════════════════════════════════════════════ | COMMENT |
| MEDIUM | recipes/tttd/examples/tttd/harness/search.py | 378 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | recipes/tttd/examples/tttd/harness/search.py | 380 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | recipes/tttd/examples/tttd/harness/search.py | 415 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | recipes/tttd/examples/tttd/harness/search.py | 417 | # --------------------------------------------------------------------------- | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | tests/reef_service/test_sao_parity.py | 42 | except ImportError: | CODE |
| HIGH | tests/reef_service/test_native_enforce.py | 404 | except RecursionError: | CODE |
| HIGH | recipes/openclawrl/examples/openclawrl/harness/agent.py | 325 | except (RuntimeError, ValueError): | CODE |
| HIGH | recipes/skillclaw/serving/judge_proxy.py | 124 | except (BrokenPipeError, ConnectionResetError): | CODE |
| HIGH | recipes/skillclaw/serving/judge_proxy.py | 160 | except (ValueError, TypeError): | CODE |
| HIGH | recipes/skillclaw/serving/judge_proxy.py | 114 | except Exception: | CODE |
| HIGH | recipes/skillclaw/harbor/tests/grade.py | 227 | except Exception: | CODE |
| HIGH | recipes/skillclaw/harbor/tests/grade.py | 351 | except Exception: | CODE |
| HIGH | recipes/skillclaw/harness/evolver.py | 517 | except (json.JSONDecodeError, ValueError): | CODE |
| HIGH | recipes/skillclaw/harness/evolver.py | 525 | except (json.JSONDecodeError, ValueError): | CODE |
| HIGH | recipes/tttd/examples/guidance_ttt/harness/state.py | 16 | except Exception: | CODE |
| HIGH | recipes/tttd/examples/tttd/harness/sandbox.py | 211 | except subprocess.TimeoutExpired: | CODE |
| HIGH | reef/core/config.py | 134 | except (ValueError, TypeError, yaml.YAMLError): | CODE |
| HIGH | reef/core/tasks/harbor.py | 444 | except ValueError: | CODE |
| HIGH | reef/runtime/executor/placement.py | 206 | except ValueError: | CODE |
| HIGH | reef/runtime/executor/process_guard.py | 103 | except ProcessLookupError: | CODE |
| HIGH | reef/runtime/executor/multiproc.py | 76 | except (EOFError, BrokenPipeError): | CODE |
| HIGH | reef/harness/tree/mutations.py | 121 | except LookupError: | CODE |
| HIGH | reef/scenario/releases.py | 164 | except ArtifactNotFound: | CODE |
| HIGH | reef/inference/process.py | 135 | except requests.RequestException: | CODE |
| HIGH | reef/artifact/git_lfs.py | 644 | except OSError: | CODE |
| HIGH | reef/service/deploy/process.py | 192 | except ProcessLookupError: | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | docker/patch/mcore_parse_hybrid_pattern.py | 1 | # ruff: noqa: F821 | COMMENT |
| LOW | tests/reef_service/test_requests_extension.py | 61 | COMMENT | |
| LOW | docs/site/lib/docs.ts | 241 | // One line of HTML: the RST generator re-indents every literal newline it | COMMENT |
| LOW | …ipes/beta/coral/examples/coral_demo/smoke/run_smoke.sh | 1 | #!/bin/bash | COMMENT |
| LOW | recipes/sdpo/examples/sciknoweval/serve.yaml | 1 | # SciKnowEval Chemistry through Reef with the sdpo recipe, for run.py | COMMENT |
| LOW | recipes/gepa/archive.py | 81 | #: Standalone archives write after every mutation. Reef-bound archives | COMMENT |
| LOW | recipes/gepa/examples/aime/run.sh | 21 | # reports every score, and seals the two test passes under ./work. The | COMMENT |
| LOW | recipes/gepa/examples/aime/gepa.yaml | 1 | # The recipe the AIME driver boots: run.py loads this file, fills | COMMENT |
| LOW | recipes/gepa/examples/aime/gepa.yaml | 21 | # The task side, in the harness/ package: evaluate is the quickstart's | COMMENT |
| LOW | recipes/sdft/examples/skill_stream/serve.yaml | 1 | # The skill stream's stack, for run.py (recipes/sdft/examples/skill_stream). | COMMENT |
| LOW | recipes/openclawrl/slime/objective.py | 1 | # Verbatim port of Gen-Verse/OpenClaw-RL's paper objective (Eq. 1): | COMMENT |
| LOW | recipes/openclawrl/slime/objective.py | 281 | # old_actor forward, NOT from rollout"), and with one optimizer step | COMMENT |
| LOW | recipes/openclawrl/slime/objective.py | 421 | # --------------------------------------------------------------------------- | COMMENT |
| LOW | recipes/openclawrl/examples/openclawrl/serve.yaml | 1 | # OpenClaw-RL on reef — next-state binary reward, changes model weights. | COMMENT |
| LOW | recipes/openclawrl/examples/openclawrl/harness/agent.py | 101 | # Sessions then burn every turn on failed patches, the student's style | COMMENT |
| LOW | recipes/basic/external-provider.yaml | 1 | # External provider inference — no GPU, no local model. | COMMENT |
| LOW | recipes/basic/local-sglang.yaml | 1 | # Local inference (SGLang) — 1+ GPUs, recipe, no training. | COMMENT |
| LOW | recipes/skillclaw/skillclaw.yaml | 1 | # The recipe the campaign driver boots: run.py loads this file, builds | COMMENT |
| LOW | …tttd/examples/guidance_ttt/solutions/lasso_qwen3_8b.py | 1 | # EVOLVE-BLOCK-START | COMMENT |
| LOW | recipes/sao/slime/__init__.py | 41 | advantages = "forbidden" | COMMENT |
| LOW | recipes/sao/examples/ceobench/serve.yaml | 1 | # SAO on CEO-Bench, for recipes/sao/examples/ceobench/run.sh. | COMMENT |
| LOW | recipes/sao/examples/ceobench/serve.yaml | 21 | # materializes a sequence's full-vocabulary logits (248k entries) in fp32 on | COMMENT |
| LOW | recipes/sao/examples/ceobench/run.sh | 21 | export CEOBENCH_TRAIN_MAX_TOKENS="${CEOBENCH_TRAIN_MAX_TOKENS:-24576}" | COMMENT |
| LOW | recipes/sao/examples/ceobench/run.sh | 41 | # bounded (reef.patch reads it) so a stuck call fails instead of stalling | COMMENT |
| LOW | recipes/sao/examples/imo_answerbench/serve.yaml | 1 | # SAO functional-smoke config for recipes/sao/examples/imo_answerbench/run.sh. | COMMENT |
| LOW | recipes/sao/examples/imo_answerbench/serve.yaml | 61 | # GAE) and --custom-pg-loss-function-path (hard-masked DIS calibration) | COMMENT |
| LOW | .github/workflows/docker.yml | 41 | # build: | COMMENT |
| LOW | .github/workflows/ci.yml | 161 | # Reuse downloads after workflow/dependency edits; uv still resolves | COMMENT |
| LOW | .github/workflows/ci.yml | 181 | # reef-client is a separate distribution (Human-Agent-Society/reef-client); | COMMENT |
| LOW | reef/runtime/recovery.py | 821 | if config.save_hf_template is not None and status != "REJECTED" and not (config.lora and marker is None): | COMMENT |
| LOW | reef/harness/adapters/descriptor.py | 301 | renderer: type[AdapterRenderer] = AdapterRenderer | COMMENT |
| LOW | reef/harness/adapters/descriptor.py | 321 | #: then makes its root under ``~/.reef/episodes``; on other platforms the root stays in the temp directory. | COMMENT |
| LOW | reef/harness/adapters/terminus/descriptor.yaml | 1 | # Terminal-Bench's Terminus 2, driven through a Reef-owned runner. | COMMENT |
| LOW | reef/harness/adapters/claude/descriptor.yaml | 1 | # Claude Code, driven headless per episode. | COMMENT |
| LOW | reef/harness/adapters/claude/descriptor.yaml | 41 | # update and install print "Updates are disabled by your administrator" and | COMMENT |
| LOW | reef/harness/adapters/claude/descriptor.yaml | 81 | defaults: | COMMENT |
| LOW | reef/harness/adapters/codex/descriptor.yaml | 1 | # OpenAI Codex CLI, driven through its non-interactive exec mode. | COMMENT |
| LOW | reef/harness/adapters/dsh/descriptor.yaml | 1 | # DeepSeek Harness (dsh), driven headless per episode. | COMMENT |
| LOW | reef/harness/adapters/dsh/descriptor.yaml | 21 | version: "0.1.2-alpha.5" | COMMENT |
| LOW | reef/harness/adapters/pi/requests.ts | 1 | // Harness requests: the /reefine and /versions commands and the | COMMENT |
| LOW | reef/harness/adapters/pi/requests.ts | 21 | // A settled step offers its install once the session is between turns, so a | COMMENT |
| LOW | reef/harness/adapters/pi/version_check.ts | 1 | // Prompt when the pulled tree is behind the channel head. An update runs only | COMMENT |
| LOW | reef/harness/adapters/pi/descriptor.yaml | 21 | PI_OFFLINE: "1" | COMMENT |
| LOW | reef/harness/adapters/hermes/descriptor.yaml | 1 | # Hermes Agent, driven headless per episode. | COMMENT |
| LOW | reef/harness/adapters/hermes/descriptor.yaml | 61 | write_json_snapshots: true | COMMENT |
| LOW | reef/harness/adapters/opencode/descriptor.yaml | 1 | # opencode, driven headless per episode. | COMMENT |
| LOW | reef/train/cordis_backend/backend.py | 1261 | """ | COMMENT |
| LOW | reef/inference/chat.py | 1201 | # usually absent from the sampled text). OpenAI-shaped clients render | COMMENT |
| LOW | reef/service/install_script.py | 261 | # the agent's own startup update check has nothing to fetch. | COMMENT |
| LOW | tutorials/reefine/configs/deployment.yaml | 21 | propose: reef.recipe.reefine.evolution:propose | COMMENT |
| LOW | tutorials/evolve-your-harness/configs/serve.yaml | 1 | # The stack `run.sh` starts: one Reef service on port 8900, state under | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | tests/reef_service/test_policy_tensors.py | 335 | # turn 0: prompt [1, 2, 99] where 99 is generation scaffold, response [3, 4] | COMMENT |
| MEDIUM | tests/reef_service/test_parallel_service_startup.py | 225 | # Exercise orchestration without starting an actual cluster. | COMMENT |
| MEDIUM⚡ | docs/site/public/slides/index.html | 36 | <section class="page" id="p1"><svg viewBox="0 0 1920 1080" width="1920" height="1080" fill="none"><defs><marker id="ah" | CODE |
| MEDIUM | recipes/beta/spade/recipe.py | 37 | # A thinking template's generation prompt ends with a think scaffold the history drops; the assembly may | COMMENT |
| MEDIUM | reef/harness/runners/native/graph.py | 86 | #: The context window a compact stage measures against when models.json names none. | COMMENT |
| MEDIUM | reef/inference/chat.py | 1204 | # (agents, judges, user simulators) sees chain-of-thought as the | COMMENT |
| MEDIUM | reef/inference/chat.py | 1216 | # chain-of-thought-as-reply this split exists to prevent, and a | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM⚡ | reef/harness/adapters/codex/bundled_models.json | 52 | "instructions_template": "You are Codex, an agent based on GPT-6. You and the user share one workspace, and your | CODE |
| MEDIUM⚡ | reef/harness/adapters/codex/bundled_models.json | 127 | "base_instructions": "You are Codex, an agent based on GPT-6. You and the user share one workspace, and your job i | CODE |
| LOW⚡ | reef/harness/adapters/codex/bundled_models.json | 52 | "instructions_template": "You are Codex, an agent based on GPT-6. You and the user share one workspace, and your | CODE |
| LOW⚡ | reef/harness/adapters/codex/bundled_models.json | 127 | "base_instructions": "You are Codex, an agent based on GPT-6. You and the user share one workspace, and your job i | CODE |
| LOW | reef/harness/adapters/codex/bundled_models.json | 182 | "instructions_template": "You are Codex, an agent based on GPT-5. You and the user share one workspace, and your | CODE |
| LOW | reef/harness/adapters/codex/bundled_models.json | 227 | "base_instructions": "You are Codex, an agent based on GPT-5. You and the user share one workspace, and your job i | CODE |
| LOW | reef/harness/adapters/codex/bundled_models.json | 277 | "instructions_template": "You are Codex, an agent based on GPT-5. You and the user share one workspace, and your | CODE |
| LOW | reef/harness/adapters/codex/bundled_models.json | 322 | "base_instructions": "You are Codex, an agent based on GPT-5. You and the user share one workspace, and your job i | CODE |
| LOW | reef/harness/adapters/codex/bundled_models.json | 368 | "instructions_template": "You are Codex, an agent based on GPT-5. You and the user share one workspace, and your | CODE |
| LOW | reef/harness/adapters/codex/bundled_models.json | 419 | "base_instructions": "You are Codex, an agent based on GPT-5. You and the user share one workspace, and your job i | CODE |
| LOW | reef/harness/adapters/codex/bundled_models.json | 628 | "instructions_template": "You are Codex, a coding agent based on GPT-5. You and the user share one workspace, an | CODE |
| LOW | reef/harness/adapters/codex/bundled_models.json | 666 | "base_instructions": "You are Codex, a coding agent based on GPT-5. You and the user share one workspace, and your | CODE |
| LOW | reef/harness/adapters/codex/bundled_models.json | 860 | "instructions_template": "You are GPT-5.2 running in the Codex CLI, a terminal-based coding assistant. Codex CLI | CODE |
| LOW | reef/harness/adapters/codex/bundled_models.json | 897 | "base_instructions": "You are GPT-5.2 running in the Codex CLI, a terminal-based coding assistant. Codex CLI is an | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tests/test_spade_training.py | 289 | CODE | |
| LOW | tests/reef_service/test_component_trainers.py | 333 | CODE | |
| LOW | tests/reef_service/_trajectories.py | 10 | CODE | |
| LOW | tests/reef_service/test_sao_bridge.py | 25 | CODE | |
| LOW | …ta/coral/examples/coral_demo/smoke/scripted_runtime.py | 43 | CODE | |
| LOW | recipes/openclawrl/slime/objective.py | 72 | CODE | |
| LOW | recipes/tttd/examples/guidance_ttt/harness/library.py | 23 | CODE | |
| LOW | recipes/tttd/examples/tttd/harness/agent.py | 25 | CODE | |
| LOW | recipes/tttd/examples/tttd/harness/search.py | 423 | CODE | |
| LOW | recipes/tttd/examples/tttd/harness/search.py | 550 | CODE | |
| LOW | reef/storage/commits.py | 264 | CODE | |
| LOW | reef/harness/runners/native/graph.py | 222 | CODE | |
| LOW | reef/harness/runners/native/serve.py | 329 | CODE | |
| LOW | reef/harness/client/tasks.py | 214 | CODE | |
| LOW | reef/scenario/releases.py | 247 | CODE | |
| LOW | reef/train/cordis_backend/backend.py | 687 | CODE | |
| LOW | reef/train/slime_backend/reef_adapters/bridge.py | 88 | CODE | |
| LOW | reef/service/runtime.py | 77 | CODE | |
| LOW | reef/service/app.py | 42 | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| CRITICAL | reef/storage/sql_records.py | 374 | self._tables.consumed.c.agent_record_id.in_(item.references), | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | …les/tttd/harbor/circle_packing_32/environment/score.py | 73 | # Check if radii are nonnegative and not nan | STRING |
| LOW | …les/tttd/harbor/circle_packing_32/environment/score.py | 82 | # Check if circles are inside the unit square | STRING |
| LOW | …les/tttd/harbor/circle_packing_26/environment/score.py | 73 | # Check if radii are nonnegative and not nan | STRING |
| LOW | …les/tttd/harbor/circle_packing_26/environment/score.py | 82 | # Check if circles are inside the unit square | STRING |
| LOW | …ults/formal-8x64-v3-packing/packing26/best_solution.py | 27 | # Check if radii are nonnegative and not nan | COMMENT |
| LOW | …ults/formal-8x64-v3-packing/packing26/best_solution.py | 36 | # Check if circles are inside the unit square | COMMENT |
| LOW | …ults/formal-8x64-v3-packing/packing32/best_solution.py | 19 | # Check if radii are nonnegative and not nan | COMMENT |
| LOW | …ults/formal-8x64-v3-packing/packing32/best_solution.py | 28 | # Check if circles are inside the unit square | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW⚡ | tests/reef_service/test_requests_extension.py | 1862 | assert events[3] == {"kind": "input", "title": "Value for SMTP_HOST", "placeholder": ""} | CODE |
| LOW | tests/reef_service/test_requests_extension.py | 1053 | assert _of_kind(out, "input") == [{"kind": "input", "title": "When?", "placeholder": ""}] | CODE |
| LOW | tests/reef_service/test_requests_extension.py | 1821 | {"kind": "input", "title": "The phone number to text, with the country code", "placeholder": ""}, | CODE |
| LOW | tests/reef_service/test_tinker_local_engine_ray.py | 160 | api_key = os.environ["TINKER_API_KEY"] if REAL else "fake-key" | CODE |
| LOW | tests/reef_service/test_version_check.py | 544 | {"kind": "input", "title": "Your Twilio account SID", "placeholder": ""}, | CODE |
| LOW | reef/inference/sglang/config.py | 23 | if self.worker_type not in {"regular", "prefill", "decode", "encoder", "placeholder"}: | CODE |
| LOW | reef/inference/sglang/config.py | 28 | or (self.worker_type != "placeholder" and self.num_gpus % self.gpus_per_engine) | CODE |
| LOW | reef/inference/sglang/launch.py | 63 | self.all_engines: list[Any] = [] if group.worker_type == "placeholder" else [None] * count | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tests/reef_service/test_native_enforce.py | 255 | "def helper(args, workdir):\n return 1\n\nNAME = 'probe'\n", | CODE |
| LOW | tests/reef_service/test_native_serve.py | 379 | broken["config"]["code"] = "def helper(args, workdir):\n return 1\n" | CODE |
| LOW | tests/reef_service/test_native_serve.py | 695 | broken["options"]["config"]["code"] = "def helper(args, workdir):\n return 1\n" | CODE |
| LOW | tests/reef_service/test_native_serve.py | 1010 | broken["config"]["code"] = "def helper(args, workdir):\n return 1\n" | CODE |
| LOW | tests/reef_service/test_native_harness.py | 1940 | "code": "def helper(args, workdir):\n return 1\n", | CODE |
| LOW | tests/reef_service/test_native_plugins.py | 205 | code="def helper(args, workdir):\n return 1\n", | CODE |
| LOW | recipes/skillclaw/harness/day.py | 559 | def run_task( | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | reef/scenario/registry.py | 245 | # Creating an existing scenario never overwrites its configuration. | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW⚡ | tests/reef_service/test_failure_manifest.py | 213 | # Step 1: both sides fail at launch, the step is rejected, and the | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW⚡ | tests/reef_service/test_failure_manifest.py | 213 | # Step 1: both sides fail at launch, the step is rejected, and the | COMMENT |