plan-dependency-provenance
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill implements strict guardrails against unauthorized code execution by explicitly forbidding the use of installation commands during the audit process.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external data from dependency manifests (e.g., package.json, requirements.txt) and lockfiles. While this presents an attack surface for indirect prompt injection, the skill's primary purpose is to audit this data for malicious patterns, and it lacks dangerous execution capabilities, minimizing the risk.
- Ingestion points: Project manifest files and lockfiles (SKILL.md).
- Boundary markers: The skill instructions emphasize a Senior supply-chain engineer role and explicit verification steps, though no specific programmatic delimiters are defined for the input data.
- Capability inventory: The skill is restricted to file writing (emitting a plan file) and implied network lookups to public registries for package verification.
- Sanitization: The skill relies on 'Guilty until verified' logic and manual verification rather than automated string sanitization.
Audit Metadata