The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to interact with the local development environment using standard tools like git and the GitHub CLI (gh). It uses these tools to perform research on project structure and to manage branch/PR workflows if available.
[PROMPT_INJECTION]: The skill's architecture for generating plans involves ingesting untrusted project data, which presents a surface for indirect prompt injection.
Ingestion points: The skill reads project structure, existing plans, and memory files to gather context (SKILL.md).
Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore potentially malicious embedded instructions in the processed files.
Capability inventory: The skill has the capability to execute git and gh commands and write Markdown files to the plans/ directory.
Sanitization: No sanitization or validation logic is defined for the content extracted from project files.
[SAFE]: The skill claims to be composed entirely of Markdown files with no executable scripts, hooks, or build steps. Analysis confirms that the logic is purely instructional and relies on the agent's native capabilities and well-known developer CLI tools.

blueprint