Skip to content

davepoon/buildwithclaude / plugins/all-skills/skills/feedback

feedback

Verified

Turn post-task feedback—even blunt or angry feedback—into an evidence-based, approval-gated improvement plan for the agent/user workflow. Use after agent work goes wrong, misses expectations, creates friction, or should improve in future. Never mutate during the feedback turn.

FileKindSize
LICENSEother1.0 KB
SKILL.mdskill md8.0 KB
references/approval-protocol.mdmarkdown1.4 KB
references/hosts/claude-code.mdmarkdown1.2 KB
references/hosts/codex.mdmarkdown1.2 KB
references/hosts/generic.mdmarkdown616 B
references/hosts/hermes.mdmarkdown1015 B
references/hosts/pi.mdmarkdown1.1 KB
references/privacy.mdmarkdown1.1 KB
references/routing.mdmarkdown2.3 KB

references/routing.md

View on GitHub
# Improvement routing

Choose the narrowest surface that can prevent recurrence.

| Destination | Use when | Avoid when |
|---|---|---|
| Runtime behavior | The correction matters only now | It should survive the session |
| Skill | A reusable method, trigger, pitfall, or verification recipe changed | It is only a preference or project fact |
| Project instructions | A repository/project rule should guide future work | It applies only to one temporary task |
| User instructions | A stable cross-project preference or collaboration contract was learned | It exposes private task detail or will go stale |
| Durable memory | The host has verified memory and the fact is stable, compact, and expected to be remembered | Memory is generated state, the lesson is procedural, or the surface is unverified |
| Project documentation | Architecture, decisions, runbooks, or operational truth changed | It is only agent behavior guidance |
| Task/issue | Work remains and belongs in the normal tracker | The proposal is not approved |
| No durable change | The incident is local, covered, or low-signal | A repeated costly failure has no guard |

## Selection test

1. Will this still matter in a month? If no, keep it runtime-only.
2. Is it a fact, preference, procedure, or project decision? Route by type.
3. Who should inherit it: one task, repository, agent, or all work?
4. Does an existing surface already own it? Revise/consolidate instead of duplicating.
5. Can it be verified? If not, narrow the claim or mark it experimental.
6. Does the permanent context cost earn its place?

Prefer correcting the responsible source, then consolidating an existing rule, then one scoped instruction, then a skill procedure. Use durable memory only for stable facts/preferences on verified hosts. Create a new system only when no existing owner fits.

## User-side improvements

Allowed only when ambiguity is evidenced, the agent could not choose a safe default, the suggestion is optional and low-effort, the agent's default without it is stated, and the suggestion does not excuse incomplete execution or false reporting.

Bad: “Write a clearer prompt next time.”

Better: “Agent commitment: before a visual redesign, surface two directions. Optional user leverage: name a non-negotiable reference. Default: preserve the current brand and propose two bounded variants.”
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall