Speck Feedback

Purpose

When developers or agents encounter a bug, a workaround, or a confusing Speck instruction, they often patch it locally or work around it without capturing the feedback. This skill defines the Continuous Feedback Capture Mechanism to ensure these learnings are captured and fed back upstream to the Speck methodology.

Core Process

1. Maintain Running Feedback File

• Write feedback immediately to a running file: .speck/feedback/<date>-<session>.md (using .speck/templates/feedback/template.md).
• Fill out the TL;DR, Gap / Observation, Symptom, Root Cause, Severity, and Diff sections.

2. Upstream Deduplication & Check

• Before filing a new issue, use gh to search for similar issues on the telum-ai/speck repository:

  gh issue list --repo telum-ai/speck --search "[keywords]"
  

• Guardrails:
  • Dedup First: Always search existing issues first.
  • Prefer Comment: If a similar issue exists, draft a comment to append to that issue instead of creating a new one.
  • Human Confirmation: Always present the drafted issue/comment to the user for one-tap confirmation before running gh issue create or gh issue comment.

3. Draft Comment or Issue

• If similar issue exists (e.g. #123):
  • Draft a comment:

    I encountered this issue in my workspace during [session].
    Here is my observation:
    - **Symptom**: [symptom]
    - **Root cause**: [root cause]
    - **Patch applied**: [patch]
    

• If no similar issue exists:
  • Draft a new issue using the feedback template.

Inline Capture Triggers

Always-on discipline demands that the moment a gate is worked around, a skill is ambiguous, or a Speck behavior is patched locally, you MUST trigger /speck-feedback: