specrail-diagnose-ci
Pass
Audited by Gen Agent Trust Hub on Jul 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local Python script
checks/route_gate.pyto manage the CI diagnostic lifecycle. - [PROMPT_INJECTION]: The skill processes untrusted external data from logs and pull request metadata, representing a surface for indirect prompt injection. 1. Ingestion points: failing workflow logs, job details, and PR metadata (SHA and branch). 2. Boundary markers: Explicit instructions provided to prevent bypassing tests, using stale evidence, or merging without human authorization. 3. Capability inventory: Execution of repository-local scripts and reproduction of failing command sequences. 4. Sanitization: Risks are managed through human review gates and manual verification of output.
Audit Metadata