j5ik2o-gh-pr-converge-loop

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill ingests untrusted content from external sources which could be used to influence the agent's behavior via indirect prompt injection.
  • Ingestion points: The skill reads GitHub PR review threads (via referenced skills) and CI failure logs (gh run view --log-failed) as described in the SKILL.md workflow.
  • Boundary markers: There are no explicit delimiters or boundary instructions provided in SKILL.md to separate system instructions from the untrusted data being processed.
  • Capability inventory: The skill has the ability to perform file modifications, git commit, and git push operations as part of its convergence loop.
  • Sanitization: The instructions do not define any sanitization, validation, or filtering of the content retrieved from PR comments or logs before it is used to generate fixes.
  • [COMMAND_EXECUTION]: The skill performs shell command execution using git and gh tools. While these are standard for the tool's purpose, they are used to automate repository mutations based on the analysis of potentially attacker-controlled external data (e.g., instructions hidden in CI logs).
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 08:26 AM
Security Audit — agent-trust-hub — j5ik2o-gh-pr-converge-loop