| name | feedback |
| description | Save agent-skills workflow feedback proactively. MUST trigger when the user corrects how a visual plan/recap, recap status block, orchestration, autonomy, docs-first, or review skill behaves ("recaps are too long", "always web-search the docs first", "don't stop to ask", "plans should lead with the wireframe") or confirms good output. Also trigger on explicit requests: "save this feedback", "show my feedback", "clear feedback", "agent-skills preferences", "consolidate feedback", "bake in my preferences". Feedback loaded automatically when running these skills.
|
| args | [{"name":"action","description":"Action to take: save, show, clear, or consolidate","required":false}] |
| user-invokable | true |
Agent-Skills Feedback Manager
Persist feedback about how the agent-skills workflow skills should behave across sessions.
Stored feedback is loaded when those skills run, so the user never has to repeat a preference.
Actions
Save Feedback
When the user corrects or confirms how an agent-skill behaves, save it:
echo '{"category": "<category>", "feedback": "<what the user said>", "context": "<optional context>"}' | \
python ${CLAUDE_PLUGIN_ROOT}/scripts/feedback_manager.py agent-skills save-feedback
Categories: plan_style, recap_style, status_format, orchestration, autonomy, docs_discipline, review, usage_limits, general
Examples:
- "Recaps are too long, keep them lean" →
{"category": "recap_style", "feedback": "Keep visual-recap output lean — prefer the headline wireframe + key-change diffs over exhaustive prose"}
- "Always lead plans with the wireframe" →
{"category": "plan_style", "feedback": "For UI plans, lead with the top canvas wireframe before document-body sections"}
- "Don't stop to ask, use best judgment" →
{"category": "autonomy", "feedback": "Default to plow-ahead: convert routine ambiguity into stated assumptions, stop only for true blockers"}
- "Web-search the official docs before guessing" →
{"category": "docs_discipline", "feedback": "Always read-the-damn-docs: web-search current official docs before implementing against any third-party API"}
- "Use red/green/yellow at the end of every response" →
{"category": "status_format", "feedback": "End substantive responses with a quick-recap red/yellow/green status block"}
- "Delegate heavy reading to cheap subagents" →
{"category": "orchestration", "feedback": "Reserve the expensive model for judgment; delegate research/coding/testing to cheaper subagents"}
Show Feedback
Display all stored feedback:
python ${CLAUDE_PLUGIN_ROOT}/scripts/feedback_manager.py agent-skills show-feedback
Present as a readable list grouped by category.
Clear Feedback
Clear all feedback or feedback for a specific category:
python ${CLAUDE_PLUGIN_ROOT}/scripts/feedback_manager.py agent-skills clear-feedback
python ${CLAUDE_PLUGIN_ROOT}/scripts/feedback_manager.py agent-skills clear-feedback recap_style
Consolidate Feedback
Graduate stable feedback into the actual SKILL.md files, making corrections permanent. This is a
Claude-driven operation — no script needed.
When to consolidate: When the user says "update the skills based on feedback", "consolidate
feedback", "bake in my preferences", or "graduate feedback into the skill".
Process:
- Load all stored feedback:
python ${CLAUDE_PLUGIN_ROOT}/scripts/feedback_manager.py agent-skills show-feedback
-
Read the target SKILL.md for the skill the feedback is about, under
${CLAUDE_PLUGIN_ROOT}/skills/<skill-name>/SKILL.md (e.g. skills/visual-recap/SKILL.md
for recap_style, skills/plow-ahead/SKILL.md for autonomy).
-
For each feedback entry, decide:
- Consolidate: Universal preferences, repeated corrections, rules that always apply.
- Keep as runtime feedback: Situational or context-dependent preferences.
-
Present a consolidation plan to the user:
## Consolidation Plan
**Will bake into SKILL.md** (permanent):
- [feedback] → edit [file]: [what will change]
**Will keep as runtime feedback** (situational):
- [feedback] → reason: [why it stays runtime]
Proceed?
-
On approval:
- Edit the target SKILL.md files using the Edit tool.
- Clear only the graduated feedback entries:
python ${CLAUDE_PLUGIN_ROOT}/scripts/feedback_manager.py agent-skills clear-feedback <category>
- Keep non-graduated entries untouched.
-
Report what changed and what remains as runtime feedback.
How Feedback is Used
When an agent-skill runs, load stored feedback and apply the matching category:
- plan_style → shapes visual-plan surface choice, layout, and what leads the document.
- recap_style → calibrates visual-recap length, wireframe choices, and diff selection.
- status_format → adjusts the quick-recap status-block convention.
- orchestration → tunes efficient-fable / efficient-frontier delegation strategy.
- autonomy → adjusts plow-ahead's stop-vs-assume threshold.
- docs_discipline → reinforces read-the-damn-docs triggers.
- review → shapes agent-watchdog / plan-arbiter scope and verdict style.
- usage_limits → calibrates stay-within-limits pause/resume behavior.
- general → applied across all agent-skills.
This ensures the user never has to repeat the same workflow preference twice.