The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements an ingestion point for untrusted data by reading a shared Team-Update-Feed from localhost:8765 via GET /api/updates. This feed contains findings and data generated by other agents working in parallel. The skill lacks instructions for boundary markers or sanitization of this external content before it is used to influence the agent's work loop. The skill possesses high-privilege capabilities, including controlling a browser via Chrome-MCP and writing to live customer data, which increases the potential impact of instructions embedded in the shared feed.
[COMMAND_EXECUTION]: The skill acts as a wrapper that orchestrates the execution of other skills (processflow-process and processflow-critic). It also interacts with a local helper application and services on localhost:8765. While this is part of its core functionality, it represents a wide execution surface across multiple tools and data stores.

processflow-run