review-extraction

Pass

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/build_golden_slice_bundle.py uses subprocess.run() to execute git diff and git --no-index. This is used exclusively for generating code diffs for the 'optional_code' review bundle when tooling changes, which is a standard development workflow within this skill's context.
  • [SAFE]: The skill implements a robust 'refusal contract' for sign-offs. In scripts/build_golden_slice_bundle.py, the _signoff_refusal_reasons function enforces deterministic quality gates, refusing to generate closure artifacts if extraction results do not perfectly match human-labeled ground truth or contain unresolved blockers.
  • [SAFE]: The skill explicitly defines 'Non-goals' in SKILL.md, prohibiting the agent from mutating core extractor code, document presets, or promotional ledgers during the evidence gathering phase. This enforces a clear separation between observation and mutation.
  • [SAFE]: The script scripts/build_golden_slice_bundle.py performs local file system operations (shutil, Path) and PDF processing (pypdfium2) within a specified output directory, adhering to the principle of least privilege for the intended task of artifact generation.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 26, 2026, 06:00 PM
Security Audit — agent-trust-hub — review-extraction