The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to download and execute a script directly from an unverified remote URL using curl -fsSL https://raw.githubusercontent.com/salomondiei08/oh-my-hermes/main/install.sh | bash. This allows for arbitrary code execution from a third-party source.\n- [COMMAND_EXECUTION]: The setup process involves cloning a third-party repository and running multiple bash scripts (install.sh, bootstrap.sh, verify.sh), which gives the skill full control over the local execution environment.\n- [PERSISTENCE]: The skill configures automated tasks in ~/.hermes/cron.d/, such as hourly issue triage and weekly security audits, establishing persistence on the host machine.\n- [CREDENTIALS_UNSAFE]: The instructions guide the user to provide and store highly sensitive API tokens for GitHub, Vercel, Supabase, and notification services in a configuration file (.hermes/config.env).\n- [INDIRECT_PROMPT_INJECTION]: The autonomous 'CTO Loop' is susceptible to indirect prompt injection as it processes data from external sources.\n
Ingestion points: The skill retrieves untrusted content from GitHub issues and pull request diffs for automated processing.\n
Boundary markers: No explicit boundaries or warnings are defined to prevent the agent from following instructions embedded within external issue descriptions or PR content.\n
Capability inventory: The skill has access to shell execution (bash), network requests, and file system modifications.\n
Sanitization: There is no description of content validation or sanitization for the data retrieved from GitHub.\n- [DYNAMIC_EXECUTION]: The skill includes a meta-skill (create-skill) that generates and saves new executable skill templates to the file system at runtime.

oh-my-hermes-workflow