The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The script scripts/warp-grep.py uses bunx to download and execute the @morphllm/morphmcp package from the NPM registry if a local server is not already configured. This occurs during the do_search function execution.
[REMOTE_CODE_EXECUTION]: The script executes remote code by calling subprocess.run on a command string that includes bunx @morphllm/morphmcp@latest. This allows for the execution of arbitrary code fetched from a remote repository at runtime.
[COMMAND_EXECUTION]: The skill relies on subprocess.run to execute system commands, specifically mcporter and bun. While it uses argument lists to prevent shell injection, it facilitates the execution of complex CLI tools with environment variables containing sensitive information (MORPH_API_KEY).
[INDIRECT_PROMPT_INJECTION]: The skill processes arbitrary codebases which serves as an untrusted data ingestion point.
Ingestion points: Files within the provided repo_path are read by the underlying warpgrep subagent.
Boundary markers: None are present to prevent the agent from obeying instructions embedded in code comments or documentation within the searched repository.
Capability inventory: Uses subprocess.run to call external MCP tools.
Sanitization: The natural language query is passed directly to the search tool without sanitization or escaping.

warp-grep