The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the copilot CLI tool via shell commands. It provides clear guidance and code examples to prevent shell injection, such as using quoted variables and avoiding the interpolation of untrusted input directly into shell source.- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from local repository files and incorporates responses from an external AI service.
Ingestion points: Local file content read via sed in SKILL.md.
Boundary markers: Use of shell quoting and formatted strings to separate instructions from data.
Capability inventory: Shell execution of the copilot binary and file reading via sed.
Sanitization: The instructions explicitly require the agent to obtain user consent before sending data and recommend redacting sensitive information.

ask-copilot