The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests untrusted data from the repository to inform development artifacts.\n
Ingestion points: The agent is instructed to read repository files, commit history, and exploration artifacts in references/phase-1-context.md.\n
Boundary markers: There are no explicit instructions for the agent to treat data as distinct from instructions or to ignore embedded commands.\n
Capability inventory: The skill enables the agent to propose and apply repository changes and execute shell commands for verification as per references/phase-5-execute.md.\n
Sanitization: Content read from external sources is not explicitly sanitized before being incorporated into the prompt context.\n- [COMMAND_EXECUTION]: The implementation phase requires the execution of shell commands to verify tasks through test-driven development (TDD).\n
Evidence: Both references/phase-5-execute.md and assets/plan-template.md detail the requirement for running tests and verification commands.\n- [SAFE]: The skill demonstrates several high-quality security design patterns.\n
A 'pairing rhythm' ensures human-in-the-loop validation for every material step, decision, and file edit.\n
Cryptographic hashing of artifacts (SHA-256) provides a robust staleness and integrity check across the SDLC phases.\n
Explicit prohibitions against unauthorized Git operations like git commit minimize the risk of unreviewed code being committed to the repository.

pair-sdlc