The Agent Skills Directory

[COMMAND_EXECUTION]: The helper script scripts/evidence-provenance.mjs spawns git and python3 processes to inspect repository state and perform file operations. This is a primary function of the skill's integrity checking system.
[REMOTE_CODE_EXECUTION]: The script scripts/evidence-provenance.mjs executes an embedded Python script string via python3 -c to provide atomic renameat2 functionality. The skill also utilizes npx to execute the gitnexus utility from the npm registry.
[EXTERNAL_DOWNLOADS]: Skill instructions describe downloading the gitnexus analyzer from the official npm registry via npx or npm install when a local runner is not present.
[PROMPT_INJECTION]: The skill ingests untrusted repository data (source code and graph metadata), creating an indirect prompt injection surface. The skill instructions contain specific 'Hard rules' to mitigate this, such as requiring source verification and forbidding the fabrication of data.

gitnexus-plan