The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Grep tool for security auditing by searching the codebase for secrets like API keys. This is a local, defensive operation to prevent sensitive data exposure.
[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it reads external configuration files to identify project technologies.
Ingestion points: Configuration files like package.json, DESCRIPTION, and Cargo.toml are read during the tech-stack identification step.
Boundary markers: None explicitly implemented for ingested file data.
Capability inventory: The agent is granted Read, Write, Edit, Bash, Grep, and Glob permissions.
Sanitization: The skill proactively instructs the agent to use placeholders for any credentials found, mitigating the risk of accidental exposure.

write-claude-md