The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/decisions.py uses the subprocess module to execute git commands. This is used to validate repository state, identify collisions with the main branch, and manage the decision record lifecycle. The script uses argument lists rather than shell interpolation, which is a safer practice.
[PERSISTENCE]: The install function in scripts/decisions.py creates a git pre-commit hook in the .git/hooks/ directory. This ensures that the check command is run automatically before any code is committed to the repository, representing a persistence mechanism in the development environment.
[PRIVILEGE_ESCALATION]: The install command applies chmod 0o755 to the pre-commit hook file to make it executable. While necessary for the hook to function, this is a modification of file system permissions.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent that the files in docs/decisions/ are "binding" instructions for the repository. This creates a surface where instructions within these markdown files could potentially influence the agent's actions if they contain adversarial content.
Ingestion points: Markdown files located in the docs/decisions/ directory.
Boundary markers: No specific delimiters or "ignore" instructions are used when the agent processes the decision record content.
Capability inventory: The agent can execute commands through scripts/decisions.py, modify files, and perform git operations.
Sanitization: The script validates the structure and metadata of records but does not filter the text content for potential prompt injection patterns.

decision-records