run-history-skill-upgrader

Pass

Audited by Gen Agent Trust Hub on Jul 9, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a bundled Python script located at scripts/validate_upgrade_artifacts.py to verify the structure and safety of modified skill packages. It also utilizes standard system tools such as py_compile and json.tool for syntax validation of modified files.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and processes untrusted data including run logs, user feedback, and external source documentation. This risk is effectively mitigated by a mandatory 'Two-Stage Approval' workflow, which prevents any file modifications until a human reviews and approves a concrete plan, and specific instructions forbidding the injection of hidden prompts into target files.
  • [SAFE]: The skill includes comprehensive safety guidelines in references/evidence-and-scope.md and SKILL.md that explicitly forbid the processing or storage of sensitive data such as API keys, passwords, tokens, or private account data. It also implements a 'Generalization Gate' to ensure that changes are based on durable patterns rather than one-off malicious or accidental incidents.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 9, 2026, 05:49 AM
Security Audit — agent-trust-hub — run-history-skill-upgrader