publish-message
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests context from external destinations such as pull request descriptions, ticket titles, and chat messages. This content is untrusted and could potentially contain malicious instructions intended to influence the agent's behavior during the drafting process.
- Ingestion points: Destination context in Step 1 (titles, descriptions, recent messages) and code diffs analyzed in 'inline-suggestions.md'.
- Boundary markers: Step 4 explicitly requires the agent to show the final text and destination to the user and wait for explicit approval before proceeding with publication.
- Capability inventory: The skill can post to multiple external services and execute CLI tools like 'gh' and 'glab'.
- Sanitization: The 'unslop.md' reference provides instructions to transform AI-generated text into plain language, though it does not provide traditional security sanitization of input data.
- [COMMAND_EXECUTION]: The skill is designed to interact with external services using verified command-line tools including the GitHub CLI ('gh') and GitLab CLI ('glab'). These operations are constrained by the skill's primary objective and the requirement for user approval.
Audit Metadata