explain-pr

Pass

Audited by Gen Agent Trust Hub on Aug 2, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes a Python script (verification_contract.py) to validate the generated PR body against a verification bundle. The command uses an absolute path resolved from the skill catalog (VERIFICATION_AUDIT_ROOT), which prevents path traversal or local file hijacking by explicitly avoiding the current working directory.
  • [DATA_EXPOSURE]: The skill processes PR metadata and verification bundles. No sensitive credentials (API keys, SSH keys, or environment secrets) are hardcoded or accessed, and no network exfiltration patterns were identified.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input such as ticket summaries and code diffs. It incorporates several security mitigations, including instructions to use literal identifiers from the bundle, a prohibition on 'stronger narratives' or reinterpreting evidence, and a mandatory validation step that compares the final output against the original verification contract.
  • Ingestion points: ticket summary, code map, diff facts, and verification bundle (SKILL.md).
  • Boundary markers: Instructions to use literal IDs and dispositions; explicit command to not add forbidden wording.
  • Capability inventory: Executes local Python script via subprocess; updates PR body via provider adapter.
  • Sanitization: Validation script compares rendered markdown against a JSON schema/contract.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 2, 2026, 10:34 AM
Security Audit — agent-trust-hub — explain-pr