The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing external data from pull request comments.\n
Ingestion points: The prd/edit-prd.md file instructs the agent to read and analyze pull request review comments and inline comments.\n
Boundary markers: No delimiters or safety instructions are provided to separate the external comments from the core prompt logic.\n
Capability inventory: The agent can perform file system writes and git operations (git push, git worktree) which could be influenced by instructions embedded within malicious comments.\n
Sanitization: There is no mention of sanitizing or validating the input from the pull request comments before the agent processes them.\n- [COMMAND_EXECUTION]: The skill uses specific git commands to manage the product documentation repository.\n
Evidence: In prd/create-prd.md, prd/edit-prd.md, and prd/reviewed-prd.md, the agent is instructed to use git fetch, git worktree, and git push to handle branch management and file operations. These are legitimate tools for the skill's purpose but represent a control surface over the local environment and remote repository.

z-product