| name | case-feedback-skill |
| description | Run the case feedback script to submit feedback for the current conversation. Only use when invoked by case-feedback-agent. |
| allowed-tools | Bash, Read |
Case Feedback Skill
Execute the feedback submission script and return the result.
Critical constraint
Run the script exactly once — regardless of success or failure, execute it once and return the outcome.
Execution
Gather information
feedback:
- If the user explicitly provided feedback text, use that directly
- If the user describes a problem or issue, summarize it concisely as the feedback
- Ask the user for clarification only if no feedback intent can be inferred
context:
The context contains a summary of the conversation, and must append the original, complete conversation history data.
Summarize the current conversation context, including:
- What task the user was trying to accomplish
- What operations were performed
- Any errors or unexpected behaviors encountered
- Relevant code snippets or file paths (keep it concise)
code_type:
Identify the programming language or code type involved (e.g., JavaScript, Python, Java). If not relevant, leave it blank.
request_id:
Extract the unique request ID or the session ID associated with this conversation or case. If not available, leave it blank.
happened_time:
Extract the timestamp when the issue occurred. If not mentioned, leave it blank.