Home / anthropics / defending-code-reference-harness
AI Slop Code RIsk Monitor

anthropics/defending-code-reference-harness

Skills for threat modeling, scanning, triage, patching, plus an autonomous scanning harness you can /customize

32.4 High pattern signal View on GitHub

Maintenance and Governance Review

This report summarizes scanner evidence for anthropics/defending-code-reference-harness, a Python project with 7,566 GitHub stars. SynthScan examined 24,645 lines of code across 138 source files, recording 519 pattern matches distributed across 12 syntactic categories. The legacy adjusted pattern signal is 32.4 (High 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.

32.4
Adjusted Signal
32.4
Raw Pattern Score
100%
Time Factor
2026-08-06
Last Push
7.6K
Stars
Python
Language
24.6K
Lines of Code
138
Files
519
Pattern Hits
2026-10-12
Scan Date
0.10
HC Hit Rate

Risk Review · v3

Use each vector to guide manual verification of the underlying findings.

Cognitive Ownership Risk 54.3/100 · HIGH

Can maintainers explain intent, assumptions, and trade-offs? 358 finding(s).

Hallucinated Dependency & Supply Chain Risk 12.4/100 · LOW

Are imports declared and package/API names verified? 46 finding(s).

Agent Over-Engineering Ratio 74.0/100 · HIGH

Are wrappers, guards, and repeated structures necessary? 437 finding(s).

PR Review Velocity Impact 80.5/100 · CRITICAL

Which changes need focused human review? 519 finding(s).

What These Metrics Mean

Adjusted Signal
Legacy pattern score normalized per 1,000 scanned lines and multiplied by the repository-level time factor. It is not an estimate of AI authorship.
Raw Pattern Score
Sum of severity- and context-weighted pattern findings before applying the time factor.
Time Factor
A coarse multiplier based on repository last-push metadata. It does not identify when individual code was written.
Pattern Hits
Count of configured pattern findings. Review each location and surrounding code before acting.
HC Hit Rate
High- and critical-severity findings per scanned file. Severity is a configured detector weight, not proof of a vulnerability or authorship.
Lines / Files
Scan coverage denominators used by the legacy normalized score.

Pattern-Signal History

This chart tracks the legacy adjusted pattern score across scan runs. Changes can reflect source changes, scan coverage, or detector/scoring versions; they do not measure AI adoption or prove that a team refactored code. Compare runs only when their scoring versions are compatible. The dashed line tracks the raw pattern-hit count.

Severity Breakdown

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.

CRITICAL 0HIGH 14MEDIUM 108LOW 397

Directory Score Breakdown

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.

Pattern Findings

The scanner identified 519 distinct pattern matches across 12 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.

Hyper-Verbose Identifiers318 hits · 314 pts
SeverityFileLineSnippetContext
LOW⚡tests/test_dnrcanary.py115def test_vuln_pipeline_refuses_dnr_target():CODE
LOW⚡tests/test_dnrcanary.py125def test_grade_passes_complete_submission():CODE
LOW⚡tests/test_dnrcanary.py133def test_grade_fails_when_idor_missed():CODE
LOW⚡tests/test_dnrcanary.py625def test_generate_logs_aborts_when_non_http_service_owns_port(tmp_path):CODE
LOW⚡tests/test_dnrcanary.py631def test_grade_cli_missing_ground_truth_exits_couldnt_grade(tmp_path):CODE
LOW⚡tests/test_dnrcanary.py641def test_committed_alerts_match_generator():CODE
LOW⚡tests/test_dnrcanary.py651def test_ground_truth_matches_source_constants():CODE
LOWtests/test_dnrcanary.py144def test_grade_idor_volume_requires_quantified_count():CODE
LOWtests/test_dnrcanary.py169def test_grade_fails_on_login_injection_claim():CODE
LOWtests/test_dnrcanary.py188def test_grade_fails_on_login_claim_with_top_level_route():CODE
LOWtests/test_dnrcanary.py208def test_grade_poc_respects_honest_negative():CODE
LOWtests/test_dnrcanary.py224def test_grade_fails_on_login_claim_in_top_level_fields():CODE
LOWtests/test_dnrcanary.py243def test_grade_passes_three_incident_split():CODE
LOWtests/test_dnrcanary.py267def test_grade_passes_ato_title_mentioning_sqli():CODE
LOWtests/test_dnrcanary.py292def test_grade_idor_scope_not_satisfied_by_ownership_mention():CODE
LOWtests/test_dnrcanary.py310def test_grade_idor_source_loc_accepts_mechanism_phrasings():CODE
LOWtests/test_dnrcanary.py332def test_grade_idor_scope_not_satisfied_by_ip_address_prose():CODE
LOWtests/test_dnrcanary.py348def test_grade_passes_ato_detail_mentioning_sqli():CODE
LOWtests/test_dnrcanary.py381def test_grade_fails_cleanly_on_non_list_fields():CODE
LOWtests/test_dnrcanary.py393def test_grade_herring_requires_burst_reference():CODE
LOWtests/test_dnrcanary.py423def test_grade_fails_cleanly_on_non_dict_submission():CODE
LOWtests/test_dnrcanary.py435def test_grade_passes_granular_sqli_split():CODE
LOWtests/test_dnrcanary.py458def test_grade_tolerates_string_entries():CODE
LOWtests/test_dnrcanary.py472def test_grade_requires_negative_finding_for_scanner():CODE
LOWtests/test_dnrcanary.py482def test_grade_rejects_herring_filed_as_suspected_incident():CODE
LOWtests/test_dnrcanary.py508def test_grade_cli_reports_invalid_json_cleanly(tmp_path):CODE
LOWtests/test_dnrcanary.py540def test_grade_cli_empty_ground_truth_exits_couldnt_grade(tmp_path):CODE
LOWtests/test_dnrcanary.py580def _assert_generate_logs_aborts(tmp_path, payload: bytes) -> None:CODE
LOWtests/test_dnrcanary.py614def test_generate_logs_aborts_when_http_server_owns_port(tmp_path):CODE
LOWtests/test_dnrcanary.py666def test_session_steps_clamp_to_window():CODE
LOWtests/test_dnrcanary.py678def test_incidents_json_example_in_readme_is_valid_json():CODE
LOWtests/test_dedup.py39def test_signature_skips_interceptor():STRING
LOW⚡tests/test_dedup.py50def test_signature_no_frame_fallback():STRING
LOW⚡tests/test_dedup.py55def test_signature_missing_fields():STRING
LOWtests/test_dedup.py105def test_dedup_groups_duplicates(tmp_path):CODE
LOWtests/test_dedup.py117def test_dedup_includes_rejected(tmp_path):CODE
LOWtests/test_dedup.py126def test_dedup_parses_operation(tmp_path):CODE
LOW⚡tests/test_dedup.py139def test_dedup_signature_prefers_parsed_type(tmp_path):CODE
LOW⚡tests/test_dedup.py149def test_dedup_skips_null_crash(tmp_path):CODE
LOW⚡tests/test_dedup.py158def test_dedup_walks_nested_batches(tmp_path):CODE
LOWtests/test_dedup.py172def test_dedup_skips_malformed_json(tmp_path):CODE
LOWtests/test_dedup.py190def test_format_report_sorted_by_count(tmp_path):CODE
LOW⚡tests/test_dedup.py205def test_format_report_shows_operation(tmp_path):CODE
LOW⚡tests/test_dedup.py212def test_format_report_shows_relative_paths(tmp_path):CODE
LOW⚡tests/test_dedup.py220def test_format_report_shows_status(tmp_path):CODE
LOWtests/test_salvage.py79def test_vp_error_without_submission_stays_agent_failed(tmp_path, monkeypatch):CODE
LOWtests/test_salvage.py87def test_vp_error_after_submission_salvages_crash(tmp_path, monkeypatch):CODE
LOWtests/test_salvage.py98def test_vp_clean_run_records_no_error(tmp_path, monkeypatch):CODE
LOWtests/test_salvage.py157def test_dnr_error_without_submission_stays_error(tmp_path, monkeypatch, capsys):CODE
LOWtests/test_salvage.py173def test_dnr_error_after_submission_salvages_and_grades(tmp_path, monkeypatch):CODE
LOWtests/test_salvage.py184def test_dnr_clean_run_records_no_hunt_error(tmp_path, monkeypatch):CODE
LOW⚡tests/test_judge.py20def test_parse_judge_dup_skip():CODE
LOW⚡tests/test_judge.py26def test_parse_judge_dup_better():CODE
LOW⚡tests/test_judge.py32def test_parse_judge_dup_without_bug_id_falls_open():CODE
LOW⚡tests/test_judge.py39def test_parse_judge_unparseable_falls_open():CODE
LOW⚡tests/test_judge.py45def test_parse_judge_case_insensitive():CODE
LOW⚡tests/test_judge.py59def test_judge_prompt_empty_manifest():CODE
LOWtests/test_judge.py73def test_judge_prompt_with_manifest():CODE
LOWtests/test_judge.py93def test_compare_prompt_has_both():CODE
LOWtests/test_judge.py113def test_manifest_append_and_read(tmp_path):CODE
258 more matches not shown…
Decorative Section Separators90 hits · 294 pts
SeverityFileLineSnippetContext
MEDIUMtargets/canary/entry.c17// ─── Parser A: heap-buffer-overflow ──────────────────────────────────────────COMMENT
MEDIUMtargets/canary/entry.c33// ─── Parser B: stack-buffer-overflow ─────────────────────────────────────────COMMENT
MEDIUMtargets/canary/entry.c45// ─── Parser C: heap-use-after-free ───────────────────────────────────────────COMMENT
MEDIUMtargets/canary/entry.c65// ─── Dispatch ────────────────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_dedup.py13# ── _signature ───────────────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_dedup.py60# ── dedup (results tree walk) ────────────────────────────────────────────────STRING
MEDIUMtests/test_dedup.py188# ── format_report ────────────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_salvage.py44# ── vuln-pipeline (harness.cli._run_once) ─────────────────────────────────────COMMENT
MEDIUMtests/test_salvage.py105# ── dnr-pipeline (dnr_harness.cli._run_once) ──────────────────────────────────COMMENT
MEDIUM⚡tests/test_judge.py11# ── judge output parsing ─────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_judge.py50# ── judge prompt building ────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_judge.py91# ── compare prompt building ──────────────────────────────────────────────────COMMENT
MEDIUMtests/test_judge.py105# ── manifest round-trip ──────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_agent.py164# ── _container_dead_reason probe ─────────────────────────────────────────────COMMENT
MEDIUMtests/test_patch_grade.py19# ── fixtures ─────────────────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_patch_grade.py49# ── PatchVerdict semantics ───────────────────────────────────────────────────COMMENT
MEDIUMtests/test_patch_grade.py130# ── T1 oracle ────────────────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_patch_grade.py149# ── ladder short-circuit (mocked docker) ─────────────────────────────────────COMMENT
MEDIUM⚡tests/test_auth.py344# ── provider ────────────────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_auth.py480# ── warn_bedrock_model ──────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_auth.py537# ── refuse_denied_auto_mode_model ───────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_auth.py622# ── required_egress_hosts ───────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_auth.py647# ── check_egress_satisfied ──────────────────────────────────────────────────COMMENT
MEDIUMtests/test_auth.py51# ── resolve_auth_env: first party ───────────────────────────────────────────COMMENT
MEDIUMtests/test_auth.py86# ── resolve_auth_env: Bedrock ───────────────────────────────────────────────COMMENT
MEDIUMtests/test_auth.py414# ── Vertex (stub) ───────────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_auth.py437# ── usage marker ────────────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_xml_parse.py11# ── parse_xml_tag ────────────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_xml_parse.py50# ── _blocks_to_text ──────────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_xml_parse.py70# ── find_tagged_message: THE BUGFIX ──────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_report.py14# ── grader tag parsing ───────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_report.py76# ── crash_file_from_frame ────────────────────────────────────────────────────STRING
MEDIUM⚡tests/test_report.py85# ── prompt building ──────────────────────────────────────────────────────────STRING
MEDIUM⚡tests/test_report.py130# ── representative picker ────────────────────────────────────────────────────STRING
MEDIUMtests/test_found_bugs.py12# ── asan.top_frame / project_frames ──────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_found_bugs.py156# ── jsonl append/read round-trip (raw ASAN excerpt format) ───────────────────STRING
MEDIUM⚡tests/test_found_bugs.py230# ── concurrent-agents prompt section ─────────────────────────────────────────STRING
MEDIUM⚡tests/test_skill_checkpoint.py346# ── input hardening ──────────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_skill_checkpoint.py36# ── happy path ───────────────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_skill_checkpoint.py104# ── rundir ───────────────────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_skill_checkpoint.py368# ── path confinement ─────────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_focus.py8# ── build_find_prompt conditional sections ───────────────────────────────────COMMENT
MEDIUM⚡tests/test_focus.py89# ── _assigned_focus round-robin ──────────────────────────────────────────────COMMENT
MEDIUMtests/test_egress_proxy.py87# ── CONNECT relay ────────────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_egress_proxy.py208# ── Tunnel idle timeout ──────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_asan.py11# ── crash_reason ─────────────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_asan.py101# ── asan_excerpt ─────────────────────────────────────────────────────────────STRING
MEDIUMtests/test_asan.py132# ── result.json integration ──────────────────────────────────────────────────STRING
MEDIUM⚡tests/test_untrusted.py35# ── helpers ──────────────────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_untrusted.py51# ── patch prompt ─────────────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_untrusted.py86# ── report prompt ────────────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_untrusted.py103# ── judge prompt ─────────────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_checkpoint.py19# ── run --resume ─────────────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_checkpoint.py57# ── --resume --stream: judge_log idempotence key ─────────────────────────────COMMENT
MEDIUMtests/test_checkpoint.py84# ── --resume layout guard ────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_checkpoint.py115# ── report (skip-on-exists) ──────────────────────────────────────────────────COMMENT
MEDIUMharness/patch_grade.py98 # ── T0: apply + build ────────────────────────────────────────────────COMMENT
MEDIUMharness/patch_grade.py146 # ── T1: PoC stops ────────────────────────────────────────────────────COMMENT
MEDIUMharness/patch_grade.py168 # ── T2: regression suite ─────────────────────────────────────────────COMMENT
MEDIUMharness/patch_grade.py186 # ── re-attack ────────────────────────────────────────────────────────COMMENT
30 more matches not shown…
AI Slop Vocabulary14 hits · 39 pts
SeverityFileLineSnippetContext
MEDIUMtargets/dnrcanary/ground_truth.yaml9# the grader scores against. The autonomous harness must never mount thisCOMMENT
MEDIUMtests/test_salvage.py44# ── vuln-pipeline (harness.cli._run_once) ─────────────────────────────────────COMMENT
MEDIUMtests/test_salvage.py140 # dnr_cli.sandbox / dnr_cli.docker_ops are the shared harness modulesCOMMENT
MEDIUMtests/test_dnr_harness.py236 # 1 — harness.cli's convention reserves 1 for usage/config/auth errorsSTRING
MEDIUMtests/test_dnr_harness.py281 # usage/config errors before any agent ran exit 1, per harness.cli'sSTRING
MEDIUMtests/test_dnr_harness.py368 # harness read other container files, e.g. the copied auth profileSTRING
MEDIUMtests/test_skill_checkpoint.py275 # skill run must NOT adopt it (which would overwrite the harness output) —COMMENT
MEDIUM⚡tests/test_skill_checkpoint.py288 assert out != str(foreign) # minted fresh, did not adopt the harness dirCODE
MEDIUM.claude/skills/_lib/checkpoint.py56# Run dirs minted by rundir: harness timestamp + optional same-second suffix.COMMENT
MEDIUMdnr_harness/cli.py430 # harness.agent_image.build captures output; without this the stderrCOMMENT
MEDIUMharness/prompts/report_prompt.py64### 2. `<reachability>` — real attack surface or harness artifact?COMMENT
MEDIUMscripts/egress_proxy.py57 # Keep in sync with harness/auth.py:_host_allowedCOMMENT
MEDIUM⚡scripts/setup_sandbox.sh93# the default from harness.auth so Bedrock/Vertex get the right host without aCOMMENT
MEDIUM⚡scripts/setup_sandbox.sh94# manual override. Needs the venv (harness package) — ensure it now.COMMENT
Unused Imports40 hits · 39 pts
SeverityFileLineSnippetContext
LOWtargets/dnrcanary/grade.py47CODE
LOWtargets/dnrcanary/generate_logs.py26CODE
LOWtargets/dnrcanary/app/seed.py10CODE
LOWtargets/dnrcanary/app/app.py13CODE
LOWtests/test_dnrcanary.py10CODE
LOWtests/test_agent.py9CODE
LOWtests/test_patch_grade.py9CODE
LOWtests/test_dnr_harness.py5CODE
LOWtests/test_patch_grade_e2e.py14CODE
LOWtests/test_patch.py5CODE
LOWtests/test_sandbox.py9CODE
LOWtests/test_untrusted.py5CODE
LOWtests/test_agent_sandbox.py12CODE
LOW.claude/skills/_lib/checkpoint.py43CODE
LOWdnr_harness/config.py11CODE
LOWdnr_harness/agent_image.py12CODE
LOWdnr_harness/system_prompt.py12CODE
LOWdnr_harness/prompts.py13CODE
LOWdnr_harness/cli.py23CODE
LOWharness/patch.py11CODE
LOWharness/config.py12CODE
LOWharness/agent_image.py12CODE
LOWharness/docker_ops.py9CODE
LOWharness/patch_grade.py17CODE
LOWharness/asan.py7CODE
LOWharness/sandbox.py13CODE
LOWharness/cli.py30CODE
LOWharness/dedup.py16CODE
LOWharness/judge.py9CODE
LOWharness/agent.py22CODE
LOWharness/find.py7CODE
LOWharness/artifacts.py7CODE
LOWharness/novelty.py13CODE
LOWharness/report.py13CODE
LOWharness/grade.py11CODE
LOWharness/recon.py11CODE
LOWharness/prompts/untrusted.py12CODE
LOWharness/prompts/system_prompt.py16CODE
LOWharness/prompts/patch_prompt.py8CODE
LOWscripts/egress_proxy.py15CODE
Empty Error Handlers7 hits · 35 pts
SeverityFileLineSnippetContext
HIGHtests/test_dnrcanary.py594except OSError:CODE
HIGHdnr_harness/cli.py292except (OSError, json.JSONDecodeError):CODE
HIGHharness/cli.py107except (FileNotFoundError, ProcessLookupError, PermissionError, IndexError):CODE
HIGHharness/agent.py229except (asyncio.TimeoutError, ProcessLookupError):CODE
HIGHscripts/egress_proxy.py124except OSError:CODE
HIGHscripts/egress_proxy.py142except OSError:CODE
HIGHscripts/egress_proxy.py132except OSError:CODE
Cross-Language Confusion6 hits · 26 pts
SeverityFileLineSnippetContext
HIGHtests/test_dedup.py31ASSERTION_OUTPUT = "entry: /work/decoder.h:4521: decode_chunk: Assertion `n >= 1 && n <= 4' failed.\n"STRING
HIGH⚡tests/test_dnr_harness.py429 # `grader:` with no value parses as YAML null; the jail must answerSTRING
HIGHtests/test_asan.py44entry: /work/img.h:1761: convert_format: Assertion `n >= 1 && n <= 4' failed.CODE
HIGHharness/patch.py89 f"({ignore}git init -q && git add -A && "CODE
HIGHharness/prompts/find_prompt.py68- SEGV at a non-null, attacker-influenced address (not 0x0 or 0x8)CODE
HIGHharness/prompts/find_prompt.py190- SEGV at a non-null, attacker-influenced addressCODE
Excessive Try-Catch Wrapping18 hits · 18 pts
SeverityFileLineSnippetContext
LOWdnr_harness/cli.py280 except Exception as e:CODE
LOWdnr_harness/cli.py362 except Exception as e:CODE
MEDIUMharness/cli.py1033def _cmd_recon(args) -> int:CODE
LOW⚡harness/cli.py667 except Exception as e:CODE
LOWharness/cli.py255 except Exception as e:CODE
LOWharness/cli.py335 except Exception as e:CODE
LOWharness/cli.py374 except Exception:CODE
LOWharness/cli.py517 except Exception as e:CODE
LOWharness/cli.py757 except Exception:CODE
LOWharness/cli.py794 except Exception as e:CODE
LOWharness/cli.py967 except Exception as e:CODE
LOWharness/cli.py1037 except Exception as e:CODE
LOWharness/cli.py1057 except Exception as e:CODE
LOWharness/cli.py1163 except Exception as e:CODE
LOWharness/cli.py1257 except Exception as e:CODE
LOWharness/cli.py1386 except Exception as e:CODE
LOWharness/agent.py402 except Exception as e:CODE
LOWscripts/setup_sandbox.sh216except Exception:CODE
Over-Commented Block10 hits · 10 pts
SeverityFileLineSnippetContext
LOWtargets/htslib/entry.c1// Copyright 2026 Anthropic PBCCOMMENT
LOWtargets/canary/entry.c1// Copyright 2026 Anthropic PBCCOMMENT
LOWtargets/drlibs/entry.c1// Copyright 2026 Anthropic PBCCOMMENT
LOWtargets/alsa/entry.c1// Copyright 2026 Anthropic PBCCOMMENT
LOWtests/__init__.py1# Copyright 2026 Anthropic PBCCOMMENT
LOWharness/auth.py361COMMENT
LOWharness/__init__.py1# Copyright 2026 Anthropic PBCCOMMENT
LOWharness/cli.py1# Copyright 2026 Anthropic PBCCOMMENT
LOWharness/prompts/__init__.py1# Copyright 2026 Anthropic PBCCOMMENT
LOWscripts/egress_proxy.py81 client = self.connectionCOMMENT
Deep Nesting8 hits · 8 pts
SeverityFileLineSnippetContext
LOWharness/patch_grade.py37CODE
LOWharness/asan.py86CODE
LOWharness/cli.py653CODE
LOWharness/agent.py87CODE
LOWharness/agent.py113CODE
LOWharness/agent.py247CODE
LOWscripts/egress_proxy.py86CODE
LOWscripts/egress_proxy.py109CODE
Modern AI Meta-Vocabulary3 hits · 7 pts
SeverityFileLineSnippetContext
MEDIUMdocs/blog-post.md187* [Claude Security](https://www.anthropic.com/product/security): Anthropic’s managed product for agentic vulnerability dCOMMENT
MEDIUMdocs/prompting.md5- [**Give the Model Room to Work**](#give-the-model-room-to-work) · High-level tasks over prescriptive scaffoldingCODE
MEDIUMdocs/prompting.md11- [**Preventing Harmful Actions**](#preventing-harmful-actions) · Prompt-level guardrails help as defense-in-depthCODE
AI Response Leakage1 hit · 5 pts
SeverityFileLineSnippetContext
HIGHdocs/blog-post.md78**Refer to the [harness README.md](https://github.com/anthropics/defending-code-reference-harness/tree/main/harness) forCODE
AI Structural Patterns4 hits · 4 pts
SeverityFileLineSnippetContext
LOWharness/patch.py30CODE
LOWharness/patch_grade.py37CODE
LOWharness/find.py21CODE
LOWharness/report.py37CODE