The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses git commands (fetch, pull, rev-parse, status, stash) to synchronize repository state and verify the environment before applying configuration changes.
[COMMAND_EXECUTION]: Utilizes filesystem utilities including find, grep, ls, and wc to perform a 'shadow check' across the directory tree, ensuring that instructions are not silently ignored by the agent due to file precedence rules.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from local repository files (e.g., package.json, pyproject.toml, Makefile, and existing .cursorrules) to generate its output. While this constitutes an indirect prompt injection surface, it is necessary for the skill's primary function of documenting project-specific commands and constraints.
Ingestion points: Project manifests, CI workflows, and existing agent configuration files (Step 2 of SKILL.md).
Boundary markers: The skill does not currently employ explicit delimiters or instructions for the agent to ignore embedded commands within the ingested data.
Capability inventory: File system write access to AGENTS.md and CLAUDE.md, and execution of local development tools via git.
Sanitization: No specific sanitization or escaping of the parsed manifest content is documented.
[EXTERNAL_DOWNLOADS]: Performs git fetch and git pull operations. These network requests are restricted to the user's pre-configured repository remote (origin) and are used to ensure consistency before automated edits.

agent-config