build
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from an external issue tracker to drive agent behavior. Ingestion points: Ticket bodies and specifications are read from the issue tracker and passed to subagents in SKILL.md and references/briefs.md. Boundary markers: The skill uses structured briefs (Builder, Fix, Verifier) to isolate data, though it does not explicitly sanitize the ticket content before interpolation. Capability inventory: The skill can execute shell commands (pnpm, git), run tests, and modify files via commits in the local repository. Sanitization: The skill employs session isolation (fresh subagents) and explicitly forbids reading sensitive configuration files like .env or .mcp.json to limit the impact of potential injections.
- [COMMAND_EXECUTION]: The skill executes shell commands and local scripts as part of its core functionality. Evidence: The skill runs check-locked-tests.sh and project-specific commands like pnpm test and pnpm lint to verify code changes.
Audit Metadata