The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute shell commands to start local application processes in the background (e.g., using uv run). This is the intended functionality for local development and testing of worker processes.
[INDIRECT_PROMPT_INJECTION]: The skill processes output from remote Diagrid Catalyst tools, creating an attack surface where instructions could be embedded in workflow execution data or logs.
Ingestion points: Output from catalyst_get_app, catalyst_get_workflow_run, catalyst_get_logs, and catalyst_get_metrics tools.
Boundary markers: The skill does not explicitly define markers to separate tool-generated content from agent instructions.
Capability inventory: The agent is authorized to execute shell commands and launch background processes.
Sanitization: The skill requires the agent to utilize placeholders for API tokens to prevent exposure in logs or chat outputs.
[SAFE]: The skill includes explicit security precautions, instructing the agent to handle DAPR_API_TOKEN as a secret. It forbids storing the token in environment files, logs, or chat prose and directs the agent to use placeholders when presenting commands to the user.

catalyst-develop