The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands to inspect and manage the installation environment.
Evidence includes the use of rm -rf and rm to delete legacy directories (agents/, commands/, skills/) and hook scripts in the ~/.claude directory during the auto-fix phase.
It uses ls, grep, and find to diagnose the presence of markers and legacy scripts.
[DYNAMIC_EXECUTION]: The skill uses node -e to run inline JavaScript for advanced file system operations.
These scripts perform version comparisons using localeCompare, handle directory traversal via readdirSync, and perform recursive deletions of the plugin cache with rmSync.
The logic is used to calculate version drift and clean up stale cache entries.
[EXTERNAL_DOWNLOADS]: The skill fetches updated configuration from the author's official GitHub repository.
It uses WebFetch to retrieve the latest CLAUDE.md from https://raw.githubusercontent.com/Yeachan-Heo/oh-my-claudecode/main/docs/CLAUDE.md.
It queries the NPM registry via npm view oh-my-claude-sisyphus version to check for available updates.
[INDIRECT_PROMPT_INJECTION]: The skill parses content from user-editable files to drive its diagnostic logic.
Ingestion points: Reads content from CLAUDE.md and companion files like CLAUDE-omc.md.
Boundary markers: No explicit boundary markers or 'ignore' instructions are used when reading these files.
Capability inventory: The skill has high-authority capabilities including file deletion (rm), remote fetching (WebFetch), and dynamic script execution (node -e).
Sanitization: The skill uses regex patterns (e.g., //i) to extract specific markers, which limits the risk of processing arbitrary instructions from the file body.

omc-doctor