The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided requirement files and documentation to drive code implementation and system changes.
Ingestion points: Reads requirements from requirements.yaml and documents within the docs/ directory.
Boundary markers: None present; the agent is instructed to directly interpret prose from external files as implementation guidelines.
Capability inventory: The agent is authorized to use git grep, write and modify code files, execute tests, and perform git commits.
Sanitization: No sanitization or filtering of requirement content is specified before the agent acts on it.
[DYNAMIC_EXECUTION]: The skill workflow involves generating and executing new code and test suites based on natural language descriptions.
Evidence: Workflow steps 2 and 3 require the agent to write new tests and implement capabilities until those tests pass, involving the creation and execution of dynamic logic (SKILL.md).

requirements-implement