The Agent Skills Directory

[SAFE]: The skill is intended for developer workflow automation and follows standard practices for repository analysis. It includes instructions to wait for user confirmation before designing or implementing tasks, which maintains a human-in-the-loop safety boundary.
[COMMAND_EXECUTION]: The instructions direct the agent to use localized shell commands, such as git log --oneline -n 5, git branch, and ls prds/*.md, to identify the current work context and target PRD file. These commands are non-privileged and appropriate for a software development tool.
[PROMPT_INJECTION]: The skill processes data from external files in the prds/ directory and git history. 1. Ingestion points: PRD markdown files, git commit messages, and branch names. 2. Boundary markers: The process uses distinct logical steps to transition from document analysis to task recommendation. 3. Capability inventory: The agent has access to bash, ls, grep, and glob for file analysis. 4. Sanitization: The agent is tasked with analytical extraction of requirements rather than direct execution of document content. This constitutes a standard document-processing surface typical for developer-focused AI tools.

prd-next