The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a Python script (scripts/inspect_repository.py) that performs local Git inspections by executing commands via subprocess.Popen. The implementation uses security best practices, such as passing arguments as a list to prevent shell injection and scrubbing sensitive Git-related environment variables (e.g., GIT_CONFIG_PARAMETERS, GIT_DIR) before execution to prevent credential leakage.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content by reading repository documentation (e.g., CONTRIBUTING.md, CLAUDE.md) and configuration files to determine project rules. Mandatory Evidence Chain: 1) Ingestion Points: Repository-local documentation and JSON configuration files (specified in references/contract.md). 2) Boundary Markers: Instructions in SKILL.md requiring the agent to distinguish facts from suggestions. 3) Capability Inventory: Local file access and Git command execution via a provided script. 4) Sanitization: Explicit instructions against exposing credentials or sensitive repository data.
[EXTERNAL_DOWNLOADS]: The skill references documentation and guidelines from well-known technology services and organizations, including GitHub, GitLab, and Microsoft. These references are used to provide authoritative workflow context and do not involve untrusted remote code execution. The skill also includes instructions for installing complementary skills from the same vendor using the npx utility.

git-recovery