The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches code and plugins from an untrusted GitHub repository (Yeachan-Heo/oh-my-claudecode) and an external npm package (oh-my-claude-sisyphus) that does not belong to the skill author or a trusted vendor.
[REMOTE_CODE_EXECUTION]: Use of the /plugin marketplace add command with a remote URL executes external code within the agent environment. The installation of external npm packages further enables execution of unverified scripts.
[COMMAND_EXECUTION]: The skill instructs the user to run setup scripts and global npm installations. It specifically enables a background daemon via omc wait --start, creating a persistent execution mechanism on the host system.
[CREDENTIALS_UNSAFE]: The configuration utility omc config-stop-callback encourages users to provide sensitive Discord webhooks and Telegram bot tokens as plaintext command-line arguments, exposing them to shell history and local process logging.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface through its multi-agent orchestration.
Ingestion points: User-provided task descriptions processed by the 'autopilot' and 'team' agents (SKILL.md).
Boundary markers: None identified to separate instructions from data within the orchestrated pipeline.
Capability inventory: The skill utilizes tools for Bash execution, filesystem modification (Write/Edit), and data retrieval (Read/Grep).
Sanitization: No input validation or sanitization of task content is specified.

omc