The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions involve the use of local command-line tools such as git log, ls, and glob to discover PRD files and analyze recent commit history. These operations are restricted to gathering local context and do not involve system modification or remote execution.- [PROMPT_INJECTION]: The skill is designed to process external data from PRD markdown files and git metadata. While this creates a potential surface for indirect prompt injection, the risk is mitigated by the skill's restricted capability set. Ingestion points: The agent reads markdown files from the prds/ directory and inspects git branch and commit information. Boundary markers: There are no explicit delimiters defined to separate the content of PRDs from the agent's internal instructions. Capability inventory: The skill uses read-only tools for discovery and explicitly delegates file updates to the user through a specific command, preventing automated modifications to the codebase. Sanitization: The instructions do not define methods for sanitizing or validating the content of the documents being analyzed.

prd-next