The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes basic filesystem discovery using ls to check for project manifest files (e.g., package.json, pyproject.toml, Cargo.toml, vercel.json, netlify.toml). It also runs git status to identify the current repository state. These are non-destructive, read-only operations intended solely for context gathering during the routing process.\n- [INDIRECT_PROMPT_INJECTION]: The skill takes user input to determine routing targets. While this ingest point is exposed to untrusted data, the skill uses it for keyword-based categorization rather than direct execution. It lacks high-privilege capabilities or network sinks that would make indirect injection a high-severity threat.

gate