The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Python's subprocess module to execute git commands for repository analysis and status checking.
Evidence: Found in scripts/implementation_contract.py within the git_status, git_value, and output_is_ignored_or_external functions.
Context: The skill correctly uses list-based arguments for subprocess.run and avoids shell=True, which prevents shell injection. The commands executed are limited to standard git utility operations (e.g., git status, git rev-parse, git check-ignore) appropriate for a developer tool.
[PROMPT_INJECTION]: The skill is designed to ingest and act upon plan.md files, which constitutes an indirect prompt injection surface.
Ingestion points: plan.md files are read in scripts/scaffold_implementation.py and parsed via scripts/implementation_contract.py.
Boundary markers: The skill mandates the presence of  and validates SHA-256 validation receipts () to ensure the plan was finalized by a trusted process.
Capability inventory: The skill can read/write files within the target repository and a designated scratch directory, and execute limited git commands.
Sanitization: Content parsing includes stripping fenced code blocks and validating structure against a strict implementation contract defined in references/implementation-contract.json.
Context: Extensive defensive measures, including the 'Implementation Hazards Catalog' and 'Implementation Protocols', are specifically implemented to mitigate the risks of processing third-party instructions.

implement-with-senior-dev