The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell and git commands to maintain repository state and local configuration.
Evidence includes shell snippets in SKILL.md and close-receipt.md for updating .gitignore (grep, printf, git commit) and verifying branch history (git rev-parse, git status).
These operations are restricted to the local repository environment and are consistent with the skill's purpose as a controller for build tasks.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external project files to guide its decision-making logic, creating a theoretical attack surface for indirect prompt injection.
Ingestion points: Processes content from requirements.md, docs/specs/INDEX.md, docs/agents/issue-tracker.md, and .skills/<CODE>/progress.md (specifically checking for status stamps and tracker configurations).
Boundary markers: The instructions guide the agent to look for specific line prefixes (e.g., Status:, Verified:) rather than treating the entire file as a direct prompt.
Capability inventory: Capability to write to the local file system (via git), manage worktrees, and modify project metadata.
Sanitization: Logic is focused on parsing specific keys and values (e.g., Approved, In-progress) to gate subsequent actions.

execute-common