The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input in the form of Raw Requirements (RR). Ingestion points: User-provided natural language requirements are analyzed and stored in work items (SKILL.md). Capability inventory: The agent is directed to use shell commands for repository management (git) and application verification (pnpm, playwright-cli) (SKILL.md). Sanitization: The skill incorporates mandatory Human-in-the-loop (HITL) checkpoints for requirement refinement, specification approval, and final verification, which serves as a robust boundary against automated execution of potentially malicious requirement content.
[SAFE]: The overall architecture and instructions prioritize traceability and verification. No obfuscation, data exfiltration, or persistence mechanisms were found. The skill's behavior is entirely consistent with its declared role as a development methodology assistant.

rdd