gitnexus-work
Pass
Audited by Gen Agent Trust Hub on Jul 19, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local development commands such as 'npm run build' and uses 'node' to run repository-relative tools like the gitnexus CLI and the evidence-provenance helper script.
- [DATA_EXPOSURE]: To maintain consistency and integrity, the skill reads metadata from Git administrative files (e.g., .git/index, HEAD, and logs). This information is used only to compute local SHA-256 digests and ensure the repository state matches the implementation plan.
- [DYNAMIC_EXECUTION]: The 'evidence-provenance.mjs' script contains a hardcoded Python macro that is executed via a subprocess to utilize the 'renameat2' system call for atomic file renames on Linux, preventing partial writes or race conditions.
- [INDIRECT_PROMPT_INJECTION]: The skill processes implementation plans from the 'docs/plans/' directory to drive agent implementation tasks. * Ingestion points: 'SKILL.md' Phase 1 loads plans through a descriptor-anchored receipt containing base64 bytes. * Boundary markers: The skill enforces an 'avoid' list and ignores instructions outside the specified 'implementation_context' schema. * Capability inventory: The agent can modify local files and execute repo-local build scripts. * Sanitization: The skill performs extensive path normalization and validates plan integrity against a pinned SHA-256 digest before execution.
Audit Metadata