The Agent Skills Directory

[PROMPT_INJECTION]: The skill advises the agent to follow the 'hint' and 'details' fields in CLI error envelopes as authoritative sources for fixes. This creates a surface for indirect prompt injection if an attacker-controlled input (such as a malicious workflow name or node comment) triggers a specific error that reflects the input back into the hint, potentially influencing the agent's behavior.
Ingestion points: The error.hint, error.code, and details fields from CLI JSON output processed in SKILL.md.
Boundary markers: None specified in the instructions.
Capability inventory: Execution of comfy CLI commands (server management, cloud operations) and local python3 execution.
Sanitization: No sanitization or filtering logic is described for the tool output.
[COMMAND_EXECUTION]: The skill provides numerous examples of legitimate troubleshooting commands for the comfy CLI and a static Python snippet for JWT decoding, which are consistent with the intended debugging purpose.
[SAFE]: The domain fe-pr-NNNNN.testenvs.comfy.org and other comfy.org references are vendor-controlled resources used for development environments.

comfy-debug