The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions mandate the installation of the product-agent package from PyPI, which is a third-party dependency from a non-standard source.\n
Evidence: Found in SKILL.md under Prerequisites and Error Handling sections: pip install product-agent.\n- [COMMAND_EXECUTION]: The skill runs the product-agent tool to perform autonomous tasks including code implementation and testing. This tool has broad capabilities to modify the file system and execute shell commands via the agent's environment.\n
Evidence: SKILL.md instructions specify running the command and monitoring its 9-phase pipeline: Intake, Architecture, Scaffold, Implement, Test, Integrate, Polish, Validate, Ship.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing untrusted user input to drive autonomous actions.\n
Ingestion points: App descriptions gathered from the user in SKILL.md (Step 1 under Instructions).\n
Boundary markers: Absent. User input is confirmed and then passed directly to the product-agent build engine without isolation or delimiters.\n
Capability inventory: The skill has access to file system tools (Read, Write, Edit, Grep, Glob) and shell execution via Bash for pip, python, npm, npx, and git as defined in the allowed-tools metadata in SKILL.md.\n
Sanitization: Absent. There is no evidence of filtering, escaping, or validating the user's description before it is interpolated into the execution command.

shipwright-pipeline