improve

Feedback Context

$ARGUMENTS

---

Capture plugin feedback from this session and file it as a GitHub issue.

Step 1: Reflect

Look back at the session. Identify:

• What went wrong or caused friction
• What the user had to correct or override
• What was missing that should exist
• What was misleading in agent/skill prompts

If $ARGUMENTS is provided, start from that. Otherwise, review the session for friction points.

Step 2: Filter — Generic vs Repo-Specific

This is the critical gate. Ask: "Would this feedback help a different project using these plugins?"

Generic (file it):

• Planner doesn't check for existing tests before proposing new ones
• /doc skill doesn't handle monorepos
• Verifier misses import-only changes

Repo-specific (don't file — belongs in CLAUDE.md):

• Custom auth headers, tool preferences (pnpm vs npm), project-specific test flags
• Exception: if the same repo-specific issue keeps appearing across projects, it's generic

Step 3: Distill

For each piece of generic feedback, structure it:

• Plugin/skill/agent affected — which specific component (e.g., wf:planner, util:doc, vault:auto-process)
• Current behavior — what happened
• Expected behavior — what should have happened
• Session context — what project, what task, what conditions triggered it

Group related feedback for the same plugin/agent cluster into one issue. Separate unrelated clusters into separate issues.

Step 4: File

For each issue, use:

gh issue create \
  --repo JSai23/claude-tooling \
  --label plugin-feedback \
  --title "<plugin>: <concise summary>" \
  --body "<structured feedback from step 3>"

Issue body format:

## Affected Component
<plugin>/<skill or agent> (e.g., wf/planner, util/doc)

## Current Behavior
<what happened>

## Expected Behavior
<what should have happened>

## Context
- **Project:** <what kind of project>
- **Task:** <what the user was doing>
- **Conditions:** <anything relevant about when/how this triggered>

## Notes
<any additional context, workarounds used, severity>

Step 5: Report

Show the user:

• Issue URL(s) created
• Summary of what was filed
• Any feedback that was filtered out as repo-specific (and where it should go instead)