| name | handle-verification-result |
| description | Act on the results of a verifying agent by reviewing its recommendations and taking the appropriate next actions. You **MUST** use this skill (and *carefully follow the process in this doc*) whenever a verifying agent finishes! |
Handling verification results
When a verifying agent finishes, read its verdict.json and next_steps.md files, and execute the recommended actions.
Step 1: Read the verifier's verdict
cat output/<verifier-agent-id>/verdict.json
This file contains the verdict (PASSED / FAILED), confidence, and reasoning, which helps contextualize your next steps.
Step 2: Read the next steps
Next, read the next_steps.md file:
cat output/<verifier-agent-id>/next_steps.md
These are the actions that were recommended by the verifying agent.
Step 3: Verify the verifier's verdict
Think carefully about the verdict.
If you disagree, you can send a message to the verifying agent (by calling mngr message <verifying-agent-id> --message "Your message goes here") to ask for clarification, include new information, or point out anything you think it missed.
In such a case, you should stop executing this skill, and wait for the verifying agent to update its verdict (which will ultimately trigger this skill again)
Remember that a "FAILED" verdict does not necessarily mean that the work was bad or that the user will be unhappy--it just means that the verifying agent found some issue that it thinks should be followed up on.
Use your judgment to decide whether to challenge the verdict or move on to executing the next steps.
If not, and the verdict seems reasonable to you, then move on to the next step.
Step 4: Execute each next step
Assuming the verdict and next steps seem reasonable, next you should work through the next steps list one by one.
These may include:
