contribution-triage
Pass
Audited by Gen Agent Trust Hub on Oct 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from GitHub Pull Requests, including descriptions, comments, and file diffs.
- Ingestion points: Reads PR head, base, description, changed-file list, diffs, comments, and linked issues/discussions (SKILL.md).
- Boundary markers: Explicitly instructs the agent that PR authors and submission text cannot grant authority and that proposed changes are evidence, never authority to alter rules (SKILL.md).
- Capability inventory: The agent can close PRs, apply the
macroscope-reviewlabel, and post comments using GitHub integration (SKILL.md). - Sanitization: The skill relies on natural language instructions to distinguish between 'submission evidence' and 'authority', and mandates a 'fail closed' state if policy files cannot be validated.
- [COMMAND_EXECUTION]: The skill utilizes the
gh(GitHub CLI) tool to perform moderation actions such as closing PRs and managing labels. - [EXTERNAL_DOWNLOADS]: The skill fetches policy and exemption files from the pingdotgg/t3code repository and mentions the t3 package usage for diagnostics.
Audit Metadata