The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the user to install a custom CLI tool from the author's GitHub repository using the command uv tool install git+https://github.com/yjmeqt/prd-tool.git.
[COMMAND_EXECUTION]: The agent is instructed to execute shell commands via the prd utility (e.g., prd validate, prd stats, prd format) to process and manage local files.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) due to its core functionality of processing external data.
Ingestion points: The agent reads content from PRD XML files (e.g., prd/<module>/<feature>.xml), the prd/README.md index, and various implementation specification Markdown files.
Boundary markers: No specific delimiters or instructions are used to separate trusted skill instructions from the content of the processed files, which could contain malicious directives.
Capability inventory: The agent has access to Read, Write, Edit, Glob, and Grep tools, as well as the custom prd CLI tool for file manipulation.
Sanitization: While the prd validate command ensures XML structural integrity, there is no mechanism described to sanitize or filter out natural language instructions that could influence the agent's behavior.

prd