ci-optimize
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted CI logs and tool outputs to diagnose performance issues.
- Ingestion points: CI logs, tool timing flags, and runner metrics fetched via API (references/measure.md).
- Sanitization: Recommends stripping terminal escape codes before processing.
- Capability inventory: Modifying repository configurations and executing binaries.
- Boundary markers: Absent, but the skill emphasizes a 'proof over trust' principle to verify tool claims.
- [REMOTE_CODE_EXECUTION]: The workflow involves the potential use of third-party tools and prerelease binaries for CI optimization.
- Mitigation: Mandates a subagent-led audit of any third-party tool source code before use (references/tools.md).
- Supply chain: Explicitly instructs pinning tools to specific commit hashes rather than moveable tags.
- Isolation: Prototypes and prerelease features must be executed in isolated scratch directories with human review.
Audit Metadata