The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs extensive system introspection using Bash commands such as find, grep, ls, git, and stat across the Task Management system and research project directories.
[EXTERNAL_DOWNLOADS]: Executes npx agnix, which involves the Node Package Runner fetching and executing a package from the npm registry if not present locally.
[DATA_EXPOSURE]: Accesses sensitive configuration files including ~/.claude/settings.json and claude_desktop_config.json to audit MCP server alignment and symlink integrity. The findings are consolidated into a local report file.
[COMMAND_EXECUTION]: Sub-agents perform environment reconnaissance by checking for the presence and responsiveness of CLI tools like gh, uv, jq, and gemini using which and version flags.
[DATA_EXPOSURE]: Scans research project directories to audit bibliography files (.bib) and check for the presence of MEMORY.md files, aggregating metadata for the health report.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content by reading the frontmatter and content of various skill, agent, and project files. While it lacks explicit boundary markers or sanitization, its operations are restricted to reporting and auditing, posing a minimal risk in this context.

system-audit