implement
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (specs, tickets, DESIGN.md) which creates a surface for indirect prompt injection. 1. Ingestion points: Reads specs, tickets, and repo files. 2. Boundary markers: Absent. 3. Capability inventory: File writing, committing, opening PRs, running tests. 4. Sanitization: Absent.
- [COMMAND_EXECUTION]: The skill instructs the agent to run tests and typechecking commands during the implementation phase.
Audit Metadata