microsoft/agent-lightning

18.4

Adjusted Score

18.4

Raw Score

100%

Time Factor

2026-08-28

Last Push

17.9K

Stars

Python

Language

16.0K

Lines of Code

129

Files

214

Pattern Hits

2026-08-29

Scan Date

0.10

HC Hit Rate

What These Metrics Mean

Adjusted Score: Primary synthetic code indicator. Raw score normalised per 1,000 lines of code and multiplied by the temporal discount factor. This is the definitive comparative metric — use it to rank repositories by AI authorship density.
Raw Score: The unmodified sum of all severity-weighted, context-multiplied pattern match scores before temporal discounting. Reflects the absolute signal strength independent of when the repository was last active.
Time Factor: The temporal discount multiplier (0–100%) applied to the raw score. Repositories last updated before ChatGPT's launch (Nov 2022) receive a 5% factor. Full signal is only assigned to repositories active in the post-adoption era (Jan 2024+).
Pattern Hits: Total count of individual pattern matches across all files and categories. A high hit count with a low score may indicate a very large codebase with isolated AI snippets; a low count with a high score indicates dense, concentrated AI signatures.
HC Hit Rate: High+Critical pattern hits per file, averaged across the repository. This orthogonal signal catches repositories where a few files are densely packed with high-severity AI tells — a strong indicator even when the normalised score appears moderate due to codebase size.
Lines of Code / Files: Total lines and files analysed. The scanner examines 94 file extensions. These denominators are used to normalise the score, enabling fair comparison between repositories of vastly different sizes.

Score History

This chart maps the temporal evolution of the adjusted synthetic code score across successive scan runs. An upward trajectory indicates ongoing incorporation of AI-generated code or expanding LLM-assisted scaffolding; a stable or declining trajectory may reflect active human refactoring, code removal, or the adoption of stricter authorship policies. The dashed secondary line (right axis) independently tracks total raw pattern hit count, which can diverge from the normalised score when codebase size changes significantly between scans.

Severity Breakdown

Classifies detected patterns by their diagnostic confidence and structural impact. CRITICAL patterns (coefficient 10) represent definitive synthetic signatures — hallucinated imports, explicit LLM attribution metadata — virtually never produced by human authors. HIGH (5) indicates strong structural tells such as cross-file repetition or cross-linguistic idioms. MEDIUM (2) covers recognisable conversational padding and AI-specific vocabulary. LOW (1) captures subtle indicators like tautological comments and generic boilerplate that require density to carry independent signal.

CRITICAL 1HIGH 12MEDIUM 9LOW 192

Directory Score Breakdown

This horizontal bar chart decomposes the repository's raw synthetic code score by top-level directory, allowing you to pinpoint precisely which modules or components carry the highest AI authorship density. Directories with disproportionately high scores relative to their size warrant targeted manual review: concentrated AI signatures often trace back to mass-generated configuration layers, auto-ported test suites, LLM-scaffolded boilerplate classes, or entire subsystems authored under heavy copilot assistance. Use this view to prioritise your human code-review effort.

Pattern Findings

The scanner identified 214 distinct pattern matches across 12 syntactic categories. Each entry below represents a discrete location in the source code where the engine recorded a statistically significant AI authorship indicator. 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 diagnostic confidence level (CRITICAL / HIGH / MEDIUM / LOW). The Context column identifies whether the match occurred inside executable code, an inline comment, or a string literal — comment-context matches receive a ×1.5 weight because LLMs systematically over-annotate. The ⚡ bolt icon marks clustered matches: three or more patterns within a 10-line window, each receiving an additional ×1.5 density multiplier as dense clusters constitute far stronger evidence of synthetic authorship than isolated hits.

Hyper-Verbose Identifiers92 hits · 98 pts

Severity	File	Line	Snippet	Context
LOW	tests/server/test_endpoints.py	73	def test_list_rollouts_filters_by_state_in(client: TestClient, auth_headers: dict[str, str]):	CODE
LOW	tests/server/test_endpoints.py	143	def test_enqueue_with_client_rollout_id_is_idempotent(client: TestClient, auth_headers: dict[str, str]):	CODE
LOW	tests/server/test_endpoints.py	176	def test_re_enqueue_does_not_clobber_a_running_rollout(client: TestClient, auth_headers: dict[str, str]):	CODE
LOW	tests/server/test_endpoints.py	199	def test_triplet_events_keep_last_model_request_for_duplicate_prompt(client: TestClient, auth_headers: dict[str, str]):	CODE
LOW	tests/server/test_endpoints.py	248	def test_triplet_events_do_not_dedupe_without_valid_prompt_tokens(	CODE
LOW	tests/server/test_endpoints.py	303	def test_proxy_completion_endpoint(client: TestClient, auth_headers: dict[str, str], monkeypatch):	CODE
LOW	tests/server/test_endpoints.py	343	def test_proxy_error_triplet_preserves_status(client: TestClient, auth_headers: dict[str, str], monkeypatch):	CODE
LOW	tests/server/test_endpoints.py	378	def test_proxy_admin_endpoints(client: TestClient, auth_headers: dict[str, str]):	CODE
LOW	tests/server/test_endpoints.py	416	def test_terminal_rollouts_cursor_pagination(client: TestClient, auth_headers: dict[str, str]):	CODE
LOW	tests/controller/test_k8s_reconciler.py	11	def test_build_job_spec_uses_agentlightning_labels() -> None:	CODE
LOW	tests/verl/test_rollout_level_advantage.py	58	def _compute_rollout_level_advantage(	CODE
LOW	tests/verl/test_rollout_level_advantage.py	74	def test_rollout_level_advantage_collapses_rollouts_and_broadcasts_triplets() -> None:	CODE
LOW⚡	tests/verl/test_rollout_level_advantage.py	117	def test_rollout_level_advantage_requires_rollout_ids() -> None:	CODE
LOW⚡	tests/verl/test_rollout_level_advantage.py	125	def test_rollout_level_advantage_rejects_inconsistent_rollout_rewards() -> None:	CODE
LOW⚡	tests/verl/test_rollout_level_advantage.py	132	def test_rollout_level_advantage_rejects_non_scalar_token_advantages() -> None:	CODE
LOW	tests/verl/test_agl_rollout_manager.py	55	def test_build_completed_rollout_skips_error_and_empty_model_requests() -> None:	CODE
LOW	tests/verl/test_agl_rollout_manager.py	148	def _triplet_model_request_event(prompt_token_ids: object, response_token_ids: list[int]) -> Event:	CODE
LOW	tests/verl/test_agl_rollout_manager.py	163	def test_extract_image_urls_from_messages() -> None:	CODE
LOW	tests/verl/test_agl_rollout_manager.py	187	def test_build_completed_rollout_aligns_image_urls_with_triplets() -> None:	CODE
LOW	tests/verl/test_agl_rollout_manager.py	225	def test_build_completed_rollout_aligns_images_without_valid_prompt_token_ids(prompt_token_ids: object) -> None:	CODE
LOW	tests/verl/test_agl_rollout_manager.py	252	def test_aligned_image_urls_count_mismatch_returns_none(capsys: pytest.CaptureFixture[str]) -> None:	CODE
LOW	tests/verl/test_agl_rollout_manager.py	259	def test_aligned_image_urls_text_only_rollout_returns_none_without_warning(	CODE
LOW	tests/verl/test_per_rollout_loss.py	36	def test_normalize_advantages_by_rollout() -> None:	CODE
LOW	tests/verl/test_per_rollout_loss.py	60	def test_policy_loss_matches_masked_sum() -> None:	CODE
LOW	tests/verl/test_per_rollout_loss.py	77	def test_normalize_advantages_validates_inputs() -> None:	CODE
LOW	tests/verl/test_rollout_adapter.py	144	def test_trajectory_prefix_mismatch_uploads_trace_merge_table_to_wandb(	CODE
LOW	tests/verl/test_rollout_adapter.py	271	def test_validation_uploads_all_compact_rollout_trajectories_to_wandb_zip(	CODE
LOW	tests/verl/test_rollout_adapter.py	420	def test_transition_image_rows_attach_multi_modal_inputs_and_mrope_position_ids(	CODE
LOW	tests/verl/test_rollout_adapter.py	447	def test_text_only_rows_keep_original_behavior_with_processor() -> None:	CODE
LOW	tests/verl/test_rollout_adapter.py	464	def test_trajectory_level_with_image_triplets_raises() -> None:	CODE
LOW	tests/verl/test_rollout_adapter.py	480	def test_truncated_image_prompt_falls_back_to_text_only(	CODE
LOW	tests/verl/test_rollout_adapter.py	514	def test_processor_without_vision_output_falls_back_to_text_only(	CODE
LOW	tests/verl/test_rollout_adapter.py	528	def test_is_mrope_processor_detection() -> None:	CODE
LOW	tests/verl/test_rollout_adapter.py	538	def test_load_pil_image_decodes_data_url() -> None:	CODE
LOW	tests/verl/test_rollout_adapter.py	552	def test_load_pil_image_fetches_remote_url_by_default(monkeypatch: pytest.MonkeyPatch) -> None:	CODE
LOW⚡	tests/examples/test_swe_smith_images.py	65	def test_openai_image_name_matches_job_template_convention() -> None:	CODE
LOW⚡	tests/examples/test_swe_smith_images.py	69	def test_openai_image_name_rejects_already_tagged_sources() -> None:	CODE
LOW⚡	tests/examples/test_swe_smith_images.py	74	def test_prepare_one_skips_when_openai_image_already_loaded() -> None:	CODE
LOW	tests/examples/test_swe_smith_images.py	89	def test_prepare_one_builds_openai_layer_from_local_base_without_pulling() -> None:	CODE
LOW	tests/examples/test_swe_smith_images.py	109	def test_prepare_one_retargs_existing_agl_openai_alias_without_build_or_pull() -> None:	CODE
LOW	tests/examples/test_swe_smith_images.py	125	def test_prepare_one_pulls_missing_base_only_when_allowed() -> None:	CODE
LOW	tests/examples/test_swe_smith_images.py	140	def test_prepare_one_can_require_preloaded_base_image() -> None:	CODE
LOW⚡	tests/examples/test_swe_smith_agent.py	321	def test_length_penalty_only_applies_to_solved() -> None:	CODE
LOW⚡	tests/examples/test_swe_smith_agent.py	328	def test_length_penalty_skips_validation() -> None:	CODE
LOW⚡	tests/examples/test_swe_smith_agent.py	337	def test_length_penalty_solved_within_budget_keeps_full_reward() -> None:	CODE
LOW⚡	tests/examples/test_swe_smith_agent.py	343	def test_length_penalty_solved_ramps_to_cap() -> None:	CODE
LOW⚡	tests/examples/test_swe_smith_agent.py	353	def test_length_penalty_guards_degenerate_span() -> None:	CODE
LOW⚡	tests/examples/test_swe_smith_agent.py	371	def test_prompt_length_penalty_only_applies_to_solved() -> None:	CODE
LOW⚡	tests/examples/test_swe_smith_agent.py	377	def test_prompt_length_penalty_skips_validation() -> None:	CODE
LOW⚡	tests/examples/test_swe_smith_agent.py	383	def test_prompt_length_penalty_below_soft_start_keeps_reward() -> None:	CODE
LOW⚡	tests/examples/test_swe_smith_agent.py	389	def test_prompt_length_penalty_ramps_linearly() -> None:	CODE
LOW⚡	tests/examples/test_swe_smith_agent.py	397	def test_prompt_length_penalty_saturates_at_hard_cap() -> None:	CODE
LOW⚡	tests/examples/test_swe_smith_agent.py	403	def test_prompt_length_penalty_stacks_with_turn_penalty() -> None:	CODE
LOW	tests/examples/test_swe_smith_agent.py	94	def test_generic_error_burns_turn_and_continues() -> None:	CODE
LOW	tests/examples/test_swe_smith_agent.py	108	def test_git_retry_removes_stale_index_lock_after_nonzero(monkeypatch, tmp_path) -> None:	CODE
LOW	tests/examples/test_swe_smith_agent.py	140	def test_git_retry_removes_stale_index_lock_after_timeout(monkeypatch, tmp_path) -> None:	CODE
LOW	tests/examples/test_swe_smith_agent.py	167	def test_git_retry_sleeps_before_each_checkout(monkeypatch, tmp_path) -> None:	CODE
LOW	tests/examples/test_swe_smith_agent.py	188	def test_checkout_bug_commit_uses_swesmith_official_checkout(monkeypatch) -> None:	CODE
LOW	tests/examples/test_swe_smith_agent.py	202	def test_forbidden_action_blocks_git_usage(monkeypatch) -> None:	CODE
LOW	tests/examples/test_swe_smith_agent.py	218	def test_forbidden_action_blocks_git_metadata_access(monkeypatch) -> None:	CODE
32 more matches not shown…

Cross-File Repetition9 hits · 45 pts

Severity	File	Snippet	Context
HIGH	examples/calc_x/train_calc_agent.py	load gsm8k datasets, build config, and launch verl training via agent lightning.	STRING
HIGH	examples/search_r1/train_search_r1_agent.py	load gsm8k datasets, build config, and launch verl training via agent lightning.	STRING
HIGH	examples/llm-in-sandbox/train_llm_in_sandbox.py	load gsm8k datasets, build config, and launch verl training via agent lightning.	STRING
HIGH	examples/gsm8k/train_gsm8k_agent.py	load gsm8k datasets, build config, and launch verl training via agent lightning.	STRING
HIGH	examples/search_r1/train_search_r1_agent.py	build the full omegaconf config by merging base + overrides.	STRING
HIGH	examples/science_world/train_sw_agent.py	build the full omegaconf config by merging base + overrides.	STRING
HIGH	examples/gsm8k/train_gsm8k_agent.py	build the full omegaconf config by merging base + overrides.	STRING
HIGH	examples/multimodal_qa/train_multimodal_qa.py	build the full omegaconf config by merging base + overrides.	STRING
HIGH	examples/llm-in-sandbox/train_llm_in_sandbox.py	build the full omegaconf config by merging base + overrides.	STRING

Unused Imports45 hits · 44 pts

Severity	File	Line	Context
LOW	tests/server/conftest.py	5	CODE
LOW	tests/server/test_endpoints.py	5	CODE
LOW	tests/verl/test_rollout_level_advantage.py	3	CODE
LOW	tests/verl/test_agl_rollout_manager.py	5	CODE
LOW	tests/verl/test_per_rollout_loss.py	3	CODE
LOW	tests/verl/test_rollout_adapter.py	3	CODE
LOW	tests/examples/test_swe_smith_images.py	3	CODE
LOW	tests/examples/test_swe_smith_agent.py	12	CODE
LOW	docs/macros/source_links.py	3	CODE
LOW	examples/calc_x/calc_agent.py	8	CODE
LOW	examples/swe_smith/pull_images.py	12	CODE
LOW	examples/swe_smith/train_smith_agent.py	3	CODE
LOW	examples/swe_smith/train_smith_agent_megatron.py	5	CODE
LOW	examples/swe_smith/agents/smith_agent.py	5	CODE
LOW	examples/search_r1/retrieval_server.py	3	CODE
LOW	examples/search_r1/train_search_r1_agent.py	3	CODE
LOW	examples/search_r1/agents/qa_em.py	17	CODE
LOW	examples/search_r1/agents/search_r1_agent.py	3	CODE
LOW	examples/science_world/train_sw_agent.py	22	CODE
LOW	examples/science_world/agents/sw_agent.py	28	CODE
LOW	examples/gsm8k/gsm8k_agent.py	8	CODE
LOW	examples/multimodal_qa/multimodal_qa_agent.py	8	CODE
LOW	examples/llm-in-sandbox/train_llm_in_sandbox.py	6	CODE
LOW	examples/llm-in-sandbox/agents/runner.py	6	CODE
LOW	scripts/check_headers.py	5	CODE
LOW	agentlightning/hooks.py	5	CODE
LOW	agentlightning/client.py	5	CODE
LOW	agentlightning/schemas.py	5	CODE
LOW	agentlightning/server/store.py	9	CODE
LOW	agentlightning/server/proxy.py	5	CODE
LOW	agentlightning/server/app.py	5	CODE
LOW	agentlightning/server/__main__.py	5	CODE
LOW	agentlightning/server/routes/models.py	5	CODE
LOW	agentlightning/server/routes/events.py	5	CODE
LOW	agentlightning/server/routes/proxy.py	5	CODE
LOW	agentlightning/server/routes/rollouts.py	5	CODE
LOW	agentlightning/controller/k8s_reconciler.py	12	CODE
LOW	agentlightning/controller/local_reconciler.py	5	CODE
LOW	agentlightning/controller/__main__.py	5	CODE
LOW	agentlightning/verl/agl_rollout_manager.py	5	CODE
LOW	agentlightning/verl/rollout_adapter.py	5	CODE
LOW	agentlightning/verl/per_rollout_loss.py	5	CODE
LOW	agentlightning/verl/entrypoint.py	13	CODE
LOW	agentlightning/verl/rollout_level_advantage.py	5	CODE
LOW	agentlightning/verl/trainer.py	5	CODE

Excessive Try-Catch Wrapping31 hits · 32 pts

Severity	File	Line	Snippet	Context
LOW	examples/calc_x/eval_utils.py	66	except Exception:	CODE
LOW	examples/swe_smith/pull_images.py	106	except Exception:	CODE
LOW	examples/swe_smith/pull_images.py	184	except Exception as exc:	CODE
LOW	examples/swe_smith/agents/smith_agent.py	350	except Exception as exc:	CODE
LOW	examples/swe_smith/agents/smith_agent.py	567	except Exception as exc:	CODE
LOW	examples/swe_smith/agents/smith_agent.py	834	except Exception as exc:	CODE
MEDIUM	examples/swe_smith/agents/smith_agent.py	557	def capture_patch() -> str:	CODE
LOW	examples/science_world/agents/sw_agent.py	224	except Exception as e:	CODE
LOW	examples/llm-in-sandbox/agents/runner.py	127	except Exception as exc:	CODE
LOW	agentlightning/client.py	58	except Exception as exc:	CODE
LOW	agentlightning/client.py	76	except Exception as exc:	CODE
LOW	agentlightning/controller/k8s_reconciler.py	152	except Exception:	CODE
LOW	agentlightning/controller/k8s_reconciler.py	266	except Exception as exc:	CODE
LOW	agentlightning/controller/k8s_reconciler.py	296	except Exception:	CODE
LOW	agentlightning/controller/k8s_reconciler.py	360	except Exception as exc:	CODE
LOW	agentlightning/controller/local_reconciler.py	44	except Exception:	CODE
LOW	agentlightning/controller/local_reconciler.py	121	except Exception:	CODE
LOW	agentlightning/controller/local_reconciler.py	235	except Exception as e:	CODE
LOW	agentlightning/controller/local_reconciler.py	253	except Exception:	CODE
LOW	agentlightning/controller/local_reconciler.py	281	except Exception as e:	CODE
LOW	agentlightning/verl/agl_rollout_manager.py	303	except Exception as exc:	CODE
LOW	agentlightning/verl/agl_rollout_manager.py	317	except Exception as exc:	CODE
LOW	agentlightning/verl/agl_rollout_manager.py	433	except Exception:	CODE
LOW	agentlightning/verl/agl_rollout_manager.py	444	except Exception:	CODE
LOW	agentlightning/verl/rollout_adapter.py	56	except Exception:	CODE
LOW⚡	agentlightning/verl/rollout_adapter.py	127	except Exception as exc:	CODE
LOW	agentlightning/verl/rollout_adapter.py	162	except Exception as exc:	CODE
LOW	agentlightning/verl/rollout_adapter.py	659	except Exception as exc:	CODE
LOW	agentlightning/verl/rollout_adapter.py	686	except Exception as exc:	CODE
LOW	agentlightning/verl/rollout_adapter.py	700	except Exception:	CODE
LOW	agentlightning/verl/entrypoint.py	134	except Exception as exc:	CODE

Cross-Language Confusion3 hits · 15 pts

Severity	File	Line	Snippet	Context
HIGH	tests/examples/test_swe_smith_agent.py	207	"cd /testbed && git checkout 614b134 -- pkg/x.py",	CODE
HIGH	examples/swe_smith/agents/smith_agent.py	93	commands connected with && or \|\|).	CODE
HIGH	examples/swe_smith/agents/smith_agent.py	151	- That block MUST contain EXACTLY ONE command (or several joined with && or \|\|).	CODE

Deep Nesting15 hits · 15 pts

Severity	File	Line	Context
LOW	examples/swe_smith/pull_images.py	149	CODE
LOW	examples/swe_smith/agents/smith_agent.py	715	CODE
LOW	examples/search_r1/retrieval_server.py	124	CODE
LOW	agentlightning/server/routes/events.py	100	CODE
LOW	agentlightning/controller/k8s_reconciler.py	161	CODE
LOW	agentlightning/controller/k8s_reconciler.py	300	CODE
LOW	agentlightning/controller/local_reconciler.py	129	CODE
LOW	agentlightning/verl/agl_rollout_manager.py	138	CODE
LOW	agentlightning/verl/agl_rollout_manager.py	166	CODE
LOW	agentlightning/verl/agl_rollout_manager.py	517	CODE
LOW	agentlightning/verl/agl_rollout_manager.py	573	CODE
LOW	agentlightning/verl/rollout_adapter.py	368	CODE
LOW	agentlightning/verl/trainer.py	286	CODE
LOW	agentlightning/verl/trainer.py	645	CODE
LOW	agentlightning/verl/trainer.py	298	CODE

Decorative Section Separators4 hits · 12 pts

Severity	File	Line	Snippet	Context
MEDIUM	tests/verl/test_rollout_adapter.py	327	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/verl/test_rollout_adapter.py	329	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	agentlightning/verl/rollout_adapter.py	188	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	agentlightning/verl/rollout_adapter.py	195	# ---------------------------------------------------------------------------	COMMENT

AI Slop Vocabulary4 hits · 12 pts

Severity	File	Line	Snippet	Context
MEDIUM	tests/examples/test_swe_smith_agent.py	245	# Network/package-install/test-harness-tamper routes are reward-hacking	COMMENT
MEDIUM	tests/examples/test_swe_smith_agent.py	294	# .git moved out of the worktree and harness now targets it explicitly.	COMMENT
MEDIUM	examples/swe_smith/agents/smith_agent.py	28	# harness still reaches it via --git-dir/--work-tree. See also _forbidden_action.	COMMENT
MEDIUM	examples/swe_smith/agents/smith_agent.py	71	# Writing test-harness/config files (conftest, .pth, sitecustomize) can force	COMMENT

Hallucination Indicators1 hit · 10 pts

Severity	File	Line	Snippet	Context
CRITICAL	agentlightning/verl/trainer.py	616	loss_mode = self.config.actor_rollout_ref.actor.policy_loss.get("loss_mode", "vanilla")	CODE

Modern Structural Boilerplate5 hits · 5 pts

Severity	File	Line	Snippet	Context
LOW	examples/science_world/agents/sw_agent.py	75	def _setup_logging() -> None:	CODE
LOW	agentlightning/verl/agl_rollout_manager.py	659	__all__ = [	CODE
LOW	agentlightning/verl/rollout_adapter.py	838	__all__ = ["RolloutAdapter"]	CODE
LOW	agentlightning/verl/entrypoint.py	28	__all__ = [	CODE
LOW	agentlightning/verl/dataset.py	12	__all__ = [	CODE

Structural Annotation Overuse3 hits · 4 pts

Severity	File	Line	Snippet	Context
LOW	docs/00-installation.md	7	#### Step 1: UV Sync	COMMENT
LOW	docs/00-installation.md	18	#### Step 2: Install `verl` and FlashAttention	COMMENT
LOW	docs/00-installation.md	35	#### Step 3: W&B Login	COMMENT

Over-Commented Block2 hits · 2 pts

Severity	File	Line	Snippet	Context
LOW	examples/swe_smith/requirements.txt	1	# Extra dependencies for the swe_smith example, beyond the Agent Lightning root	COMMENT
LOW	examples/search_r1/agents/qa_em.py	1	# Copyright (c) Microsoft. All rights reserved.	COMMENT

Analysis Overview

What These Metrics Mean

Score History

Severity Breakdown

Directory Score Breakdown

Pattern Findings