ecc-contribution-review

Pass

Audited by Gen Agent Trust Hub on Nov 16, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill's test suite (tests/record_security_evidence.py and tests/test_review_regressions.py) utilizes subprocess.run to execute local scripts and security scanners (Bandit). These calls use shell=False and fixed arguments, posing no risk to the host environment during normal operation of the compiler script.
  • [DYNAMIC_EXECUTION]: The test environment uses importlib to dynamically load the main compiler script for unit testing purposes. This is a standard Python development practice and does not process untrusted data for execution.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data via JSON observation files. The contribution_review_validation.py script implements rigorous schema validation, regex checks for hashes, and strict path sanitization (safe_repo_path) to mitigate injection risks. The compiler script specifically rejects non-standard JSON extensions like NaN or Infinity to ensure consistent parsing.
  • [SAFE]: The main script (compile_contribution_review.py) is designed with a restrictive posture toward the host system. It refuses to overwrite existing files using the 'x' open mode and performs no network operations, Git actions, or unauthorized external effects. The design documentation and workflow instructions emphasize that the tool only compiles claims and does not verify or execute them.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 16, 2026, 08:16 PM