setup
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill reads project-specific files like AGENTS.md and CLAUDE.md to identify existing configurations. This creates a surface where the agent could ingest untrusted data containing malicious instructions. However, the skill implements a strict multi-step workflow requiring the agent to present a concrete proposal with diffs for human review and explicit approval (Step 5) before any changes are applied, effectively mitigating the risk of unauthorized actions.\n
- Ingestion points: AGENTS.md, CLAUDE.md, and issue tracker data fetched via CLI tools (identified in SKILL.md Step 1).\n
- Boundary markers: Not explicitly defined in the file-reading instructions.\n
- Capability inventory: File system modifications and execution of gh or glab CLI commands (identified in SKILL.md Step 6).\n
- Sanitization: Relies on human-in-the-loop validation of the proposed changes before implementation.\n- [COMMAND_EXECUTION]: The skill instructs the agent to use the gh (GitHub) and glab (GitLab) command-line interfaces for project management tasks, such as managing triage labels and issues. These operations are restricted to the project's own repository and are subject to the user approval process defined in the skill's primary workflow.
Audit Metadata