The Agent Skills Directory

[COMMAND_EXECUTION]: Several scripts (init-workspace.py, route.py, update-manifest.py) utilize the subprocess module to execute local commands. Specifically, the scripts execute git rev-parse HEAD to capture the current repository state for audit trails. Additionally, update-manifest.py dynamically invokes validation scripts from sibling directories that belong to the same authorized skill suite (e.g., replacement-governance, recover-behavioral-contracts).
[INDIRECT_PROMPT_INJECTION]: The skill processes project-specific data stored in the .replacement/ directory, including manifests and evidence JSON files.
Ingestion points: The system reads manifest.json, governance.json, and various evidence artifacts from the local .replacement/ directory in the repository being analyzed.
Boundary markers: The scripts utilize internal JSON schemas and explicit allowed-value sets (such as the VALID and ALLOWED constants) to constrain the interpretation of data.
Capability inventory: The skill's capabilities are limited to local file system operations (read/write within the .replacement/ folder) and subprocess calls to Git and sibling suite scripts.
Sanitization: The skill implements mandatory integrity checks by comparing SHA256 checksums and repository revisions for all evidence artifacts before allowing state updates, which mitigates the risk of processing tampered or stale data.
[EXTERNAL_DOWNLOADS]: The documentation and error messages provide instructions for installing the complete skill suite via npx skills@latest, pointing to the author's official repository 53able/large-system-replacement-skills. This represents standard and transparent dependency management for the platform.

replace-large-system