comet-github-issue-triage

Pass

Audited by Gen Agent Trust Hub on Aug 14, 2026

Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill requires the agent to read and process external GitHub issues and comments, which introduces a surface for indirect prompt injection where an attacker could embed instructions to influence the agent's behavior.\n
  • Ingestion points: Step 1 in SKILL.md specifies reading issue bodies, comments, and related pull requests.\n
  • Boundary markers: The instructions lack specific delimiters or guidance to treat external content as untrusted, increasing the risk that the agent may follow instructions found within an issue.\n
  • Capability inventory: The agent is granted read access to the local repository and maintainer contracts (e.g., ../comet-github/references/maintainer-contract.md). However, the skill explicitly prohibits the agent from performing automated write actions such as closing issues, adding labels, or creating follow-up issues, which significantly limits the potential impact of an injection attack.\n
  • Sanitization: No sanitization or validation of the input data is mentioned in the instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 14, 2026, 02:59 PM
Security Audit — agent-trust-hub — comet-github-issue-triage