From AFK
Use when afk:implement has landed, when the user asks to simplify or clean up changed code, or when they ask to deduplicate, DRY, or review quality without hunting for bugs
How this skill is triggered — by the user, by Claude, or both
Slash command
/afk:simplifyThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Improve changed code quality by finding and fixing reuse, simplification,
Improve changed code quality by finding and fixing reuse, simplification, efficiency, and altitude issues. This is not a correctness review: do not hunt for bugs or broaden the scope beyond the reviewed changes.
Core principle: review the target diff from four independent cleanup angles, deduplicate the findings, then apply only fixes that preserve intended behavior.
Use this after afk:implement lands or when the user asks for changed code to
be simplified, cleaned up, deduplicated, or made DRY.
If the user passes a PR number, branch name, or file path, use that argument as the review target. Otherwise, review the current changed-code diff.
Gather the review scope.
Run git diff @{upstream}...HEAD to get the unified diff under review. If
there is no upstream, use git diff main...HEAD or git diff HEAD~1.
If there are uncommitted changes, or the range diff is empty, also run
git diff HEAD and include working-tree changes in scope. This review often
runs before the commit.
If the user supplied a PR number, branch name, or file path, review that target instead.
Launch four independent review agents in parallel.
Use the Task tool in a single message so all four run concurrently. Pass each agent the diff and exactly one cleanup angle, with its definition:
Reuse: Flag new code that re-implements something the codebase already has. Grep shared or utility modules and files adjacent to the change, then name the existing helper to call instead.
Simplification: Flag unnecessary complexity added by the diff: redundant or derivable state, copy-paste with slight variation, deep nesting, or dead code left behind. Name the simpler form that does the same job.
Efficiency: Flag wasted work added by the diff: redundant computation, repeated I/O, independent operations run sequentially, or blocking work added to startup or hot paths. Also flag long-lived objects built from closures or captured environments because they keep the enclosing scope alive for the object's lifetime; prefer a class or struct that copies only the fields it needs. Name the cheaper alternative.
Altitude: Check that each change is implemented at the right depth, not as a fragile bandaid. Special cases layered on shared infrastructure indicate the fix may not be deep enough; prefer generalizing the underlying mechanism over adding special cases.
Require each agent to return findings with file, line, a one-line
summary, and the concrete cost: what is duplicated, wasted, or harder to
maintain.
Aggregate and apply the fixes.
Wait for all four agents to complete. Deduplicate findings that point at the same line or mechanism. Do not re-review the diff yourself; the agents already did the review. Fix each remaining valid finding directly.
Skip findings whose fix would change intended behavior, require changes well outside the reviewed diff, or appear to be false positives. Note each skip instead of arguing with it.
STOP and ask when the intended behavior is ambiguous and the cleanup would pick between product outcomes.
STOP and ask when the only plausible fix requires changing files or systems far outside the reviewed diff.
Do not ask before skipping a finding that is clearly a false positive or clearly outside this skill's quality-only scope; note the skip in the output.
| Thought | Reality |
|---|---|
| "This looks buggy." | Correctness review is out of scope. Only fix it if the cleanup is behavior-preserving. |
| "The helper could be generalized later." | If the diff duplicated an existing helper now, use the existing helper now. |
| "The special case is small." | Small special cases on shared infrastructure may be altitude issues. Check whether the mechanism should be generalized. |
| "One agent found it, so apply it." | Deduplicate and judge findings before editing. Skip behavior changes, broad rewrites, and false positives. |
Finish with a brief summary in this shape:
Changed:
- [reuse/simplification/efficiency/altitude] What was fixed.
Skipped:
- Finding skipped and why.
Verification:
- Commands run, or "Not run" with the reason.
If the code was already clean, say that no cleanup changes were needed and include any verification run.
npx claudepluginhub alexanderop/afk --plugin afkFetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.
Whole-repo audit for over-engineering: finds dead code, unnecessary abstractions, stdlib-replaceable dependencies. Outputs ranked findings and net line/dep savings.