The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands to manage Git repositories and run project-specific verification tools like npm, make, and pytest.
[PROMPT_INJECTION]: The skill is exposed to indirect prompt injection via untrusted project files but implements strong defensive measures.
Ingestion points: Processes data from package.json, Makefile, README.md, and git diff outputs.
Boundary markers: Includes a specific 'ANTI-INJECTION MANDATE' instructing the agent to ignore instructions embedded in data.
Capability inventory: Terminal access for executing Git commands and build toolchains.
Sanitization: Implements a 'Safety Review' requiring the agent to scan commands for malicious patterns and obtain explicit user confirmation before execution.

git-commit