The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the target repository, creating a surface for indirect prompt injection.
Ingestion points: Analyzes repository structure via ls and find, reads package manifests, parses source code comments, and processes git log output (commit messages) in Phase 1.
Boundary markers: Employs a mandatory "Discovery Summary" phase that explicitly requires human confirmation before the agent proceeds to generate documentation in Phase 2.
Capability inventory: Performs file reads, shell commands for repository discovery, and git operations for branching and committing.
Sanitization: The instructions explicitly forbid the inclusion of secret values (passwords, API keys) and require that all generated claims be traceable to empirical evidence identified in the codebase.
[COMMAND_EXECUTION]: The skill executes various system and git commands to perform its documented discovery and versioning tasks.
Utilizes ls -la, find, and git log to gather repository metadata during the discovery phase.
Performs git checkout and git commit to manage the documentation lifecycle and delivery.
Incorporates the use of markdown linters such as markdownlint, remark, or vale to verify the generated documentation files.

create-readme