The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute discovery commands such as find, ls, and cat on user-provided repository paths to analyze documentation structure and content. It also executes various internal Python scripts (aggregate_benchmark.py, package_skill.py, run_loop.py) to manage the benchmarking, packaging, and optimization workflows.
[EXTERNAL_DOWNLOADS]: The interactive evaluation viewer (viewer.html) fetches the SheetJS library (xlsx.full.min.js) from cdn.sheetjs.com, a well-known technology service used for rendering spreadsheet data within the browser.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes external, potentially untrusted documentation to generate new instructions and run evaluations.
Ingestion points: The skill reads documentation files (.md, .adoc, .rst) and configuration files from local paths provided by the user in SKILL.md.
Boundary markers: While it uses YAML frontmatter and markdown sections for organization, it lacks explicit instructions to ignore or sanitize instructions embedded within the source documents.
Capability inventory: The skill possesses capabilities for shell command execution, Python script execution, and starting a local web server.
Sanitization: No explicit content filtering or escaping of documentation content is performed before it is synthesized into the generated SKILL.md file.

ing-skill-generator