The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains an inherent surface for indirect prompt injection as it processes untrusted local repository content to synthesize documentation.
Ingestion points: The evidence-gathering workflow in SKILL.md and references/module-research.md reads project files, package structures, and dependency graphs.
Boundary markers: Not explicitly defined to separate content from instructions during the synthesis phase.
Capability inventory: The skill possesses file-reading, searching (ripgrep), and file-writing (AGENTS.md) capabilities.
Sanitization: No specific content sanitization or instruction-filtering logic is documented.
Context: This behavior is fundamental to the skill's purpose of documenting existing codebases and is generally acceptable when applied to local development environments.- [COMMAND_EXECUTION]: The skill instructs the agent to utilize common command-line utilities for repository analysis and maintenance.
Evidence: Recommended tools include rg (ripgrep), sed, cat, and the repository packing tool repomix for module research, as well as mdformat and markdownlint for output validation.
Context: The usage is restricted to local repository investigation and follows best practices, such as checking for tool availability before execution.

create-agents-files