The Agent Skills Directory

[SAFE]: The skill implements a robust planning workflow using issue trackers and architectural records. It functions as an orchestration layer for development tasks and does not perform unauthorized network operations, data exfiltration, or malicious code execution.
[PROMPT_INJECTION]: The skill identifies an indirect prompt injection surface through the ingestion of external data. 1. Ingestion points: The agent reads ticket bodies, comments, and metadata from a tracker into the model context. 2. Boundary markers: The skill relies on markdown structure but does not employ explicit isolation delimiters for untrusted tracker content. 3. Capability inventory: The skill can invoke other tools for research (/research), prototyping (/prototype), and modifying project files (docs/adr/, CONTEXT.md). 4. Sanitization: No specific sanitization or filtering of the tracker-provided content is performed. This surface is inherent to the skill's primary purpose and is managed by user-in-the-loop interactions.

wayfinder