The Agent Skills Directory

[SAFE]: The skill follows secure patterns for codebase interaction. It focuses on generating documentation and requires user verification before writing files to the disk.
[COMMAND_EXECUTION]: The skill uses tools like Bash and Grep to explore the repository. These actions are scoped to the intended purpose of verifying the current state of the project for PRD accuracy.
[DATA_EXFILTRATION]: While the skill reads codebase content, it does not have access to network tools and does not attempt to transmit data to external servers.
[PROMPT_INJECTION]: The skill analyzes external data (the codebase), which represents a surface for indirect prompt injection. This is addressed by a mandatory user review step of the PRD outline before the final document is generated.
Ingestion points: Local codebase files accessed during the 'Verify' step in SKILL.md.
Boundary markers: None explicitly defined for file content, but the workflow includes a 'Confirm the outline' step (Step 5) requiring user approval.
Capability inventory: Read, Write, Edit, Bash, Glob, Grep, Skill.
Sanitization: No automated sanitization of ingested code content is specified.

create-a-prd