The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses several shell utilities to analyze the structural integrity and freshness of files. This includes awk for verifying frontmatter delimiters, grep for pattern discovery, and git log or find for determining modification history. It also contains logic to execute a local JavaScript file via node tools/pattern-alignment.js if that tool is present in the repository.
[EXTERNAL_DOWNLOADS]: The skill performs limited network operations to validate the availability of external links and dependencies cited in audited skills. It uses curl -sI to fetch HTTP headers and the gh CLI to interact with the GitHub API for repository metadata. These requests are restricted to metadata retrieval and do not execute remote code.
[PROMPT_INJECTION]: The skill processes potentially untrusted content from other skills' documentation, creating a surface for indirect prompt injection.
Ingestion points: The skill reads SKILL.md files from various subdirectories within the skills/ root.
Boundary markers: It implements 'Boundary Rules' that explicitly instruct the agent to treat the audited files as read-only data and to avoid dynamic execution.
Capability inventory: The skill has access to shell tools and network utilities for validation purposes.
Sanitization: No technical sanitization of the input text is performed; safety relies on the high-level boundary instructions and the agent's defined role as an auditor.

skill-health