click-path-audit-edho-ferdian
Trace each button, toggle or form submit through its full state changes to find handlers that undo each other, async races, stale closures and effects that reset state. Use for "tombolnya gak jalan", "diklik tapi gak ada yang terjadi", "the button does nothing", "state-nya balik lagi", or after refactoring a shared store.
- License 许可
- MIT
- 所属仓库
- edhoferdian/EEF
- 最近更新
- 2026-10-30
- Stars
- ★ 22
- Forks
- 0
01 前言 Front Matter
- name
- click-path-audit-edho-ferdian
- metadata
- { "last_reviewed": "2026-10-07" }
- description
- Trace each button, toggle or form submit through its full state changes to find handlers that undo each other, async races, stale closures and effects that reset state. Use for "tombolnya gak jalan", "diklik tapi gak ada yang terjadi", "the button does nothing", "state-nya balik lagi", or after refactoring a shared store.
03 SKILL.md 正文
Click-Path Audit — Edho Ferdian Mode
Ordinary debugging asks: does the handler exist, does it crash, are the types right. All three can be YES while the button is still broken. This skill asks the question those three miss:
Does the final state match what the control's label promises?
The defect class
A "New Email" button called setComposeMode(true) and then
selectThread(null). Both functions existed, neither threw, types were
correct — and selectThread reset composeMode: false as an undeclared side
effect. The button did nothing. A systematic debugging pass that found 54 other
defects missed this one, because nothing about it is visible from either
function in isolation. It only appears when you look at the sequence.
Scope first
This audit is expensive; scope it before starting:
| Scope | When |
|---|---|
| One control | A user reported one specific broken button |
| One screen | A new page was just built, or one screen misbehaves |
| One store | A shared store action was modified — audit every caller of the changed actions |
| Whole app | Pre-release, or after a refactor that touched shared state broadly |
For a whole-app audit, Step 1 must complete before any Step 2 work begins —
its output is the input for everything else. On a harness with sub-agent
delegation, build the map yourself, then fan out Step 2 to
click-path-tracer-edho-ferdian in parallel — one delegate per
screen/module, each given the complete map, never building its own
partial one. This is worth the delegation overhead at whole-app scale
(potentially dozens of touchpoints, each independent to trace once the
map exists) but not for the smaller scopes above — trace those inline.
On a harness with no delegation primitive, trace every touchpoint inline
regardless of scope, per Step 2 below. When a tracer's report doesn't hold
up against the map (it cites a handler or a store action that isn't
there), re-run that one screen one tier up rather than the whole fan-out —
config-hygiene-edho-ferdian's references/harness-operation.md §4.2.
Step 1 — Build the side-effect map (mandatory, always first)
For every state store in scope (Zustand store, Redux slice, React context reducer, signal container, view model), record for each action:
- which fields it sets
- which fields it resets as a side effect — fields it does not conceptually own
STORE: emailStore
setComposeMode(bool) → sets {composeMode}
selectThread(t|null) → sets {selectedThread, selectedThreadId, messages, drafts}
RESETS {composeMode: false, composeData: null, redraftOpen: false}
setDraftGenerating(b) → sets {draftGenerating}
DANGEROUS RESETS (an action clearing state it does not own):
selectThread → resets composeMode (owned by setComposeMode)
reset → resets everything
The DANGEROUS RESETS list is the whole point of this step. Every defect this skill exists to catch lives there. Do not proceed to Step 2 with an incomplete map — an audit against a partial map produces false confidence, which is worse than no audit.
Step 2 — Trace each touchpoint
For every interactive element in scope (onClick, onSubmit, onChange,
keyboard handler, gesture handler):
TOUCHPOINT: "New Email" — ThreadList.tsx:88
HANDLER: onClick
1. setComposeMode(true) → sets {composeMode: true}
2. selectThread(null) → RESETS {composeMode: false} ← CONFLICT
EXPECTED (from the label): a blank compose form opens
ACTUAL: composeMode is false; nothing renders
VERDICT: BUG — Sequential Undo
Check every trace against these six patterns:
- Sequential undo — a later call resets what an earlier call set. The canonical case above.
- Async race — two async calls both write the same field; the final value depends on resolution order, not on intent.
- Stale closure — a memoized handler captures an old value, so two increments apply the same stale base and the effect happens once, not twice.
- Missing transition — the handler validates, logs, or sets a flag but never performs the action the label promises (no API call, no navigation, no persistence).
- Conditional dead path — the real work sits behind a condition that is always false at that point in the lifecycle.
- Effect interference — the handler sets a field and a watcher/effect observing that field immediately resets it.
For each call in a trace, answer four questions: what does it read, what does it write, does it touch shared state, and does it reset anything as a side effect.
Step 3 — Report
CLICK-PATH-001 [HIGH]
Touchpoint: "New Email" — src/components/ThreadList.tsx:88
Pattern: Sequential Undo
Trace:
1. setComposeMode(true) → sets {composeMode: true}
2. selectThread(null) → RESETS {composeMode: false}
Expected: blank compose form opens
Actual: nothing renders; composeMode is false by the end of the handler
Fix: reorder (selectThread first, then setComposeMode), or remove the
undeclared composeMode reset from selectThread and clear it at the
call sites that actually mean to
Severity: CRITICAL — a money, auth, or destructive control that silently does nothing (the user believes the action happened). HIGH — a primary control that does nothing. MEDIUM — a control whose final state is partly wrong. LOW — a redundant call with no user-visible effect (dead code, not a defect).
Boundaries — what this is not for
- API-level defects (wrong response shape, missing endpoint) — a click-path
trace ends at the request; take those to normal debugging, and to
test-authoring-edho-ferdian/references/regression-testing.mdfor the test. - Styling and layout — visual inspection, or a QA sweep
(
e2e-testing-edho-ferdian/references/qa-sweep.md). - Performance —
performance-audit-edho-ferdian.
Handoff — every finding earns a test
A click-path finding is a proven defect with a known reproduction, which makes
it the ideal RED gate. Hand each one to dev-kickoff-edho-ferdian's normal
PLAN → TEST → IMPLEMENT loop: the failing test asserts the final state after
the handler runs, not that each individual function was called. A test that
asserts setComposeMode was called would have passed against the original bug.
Language routing (fixed — see skill-authoring-edho-ferdian's canonical contract)
Communication to the user in Bahasa Indonesia; the trace report and call
sequence in English, since it gets pasted into an issue or handed to
dev-kickoff-edho-ferdian's TEST/IMPLEMENT loop. Full contract:
skill-authoring-edho-ferdian §7.
Global rules
- Store map before traces, always. No exceptions; a partial map hides the exact defect this skill targets.
- Trace in execution order. The order is the evidence.
- Judge against the label, not against the code. The label is the contract with the user.
- A finding is a trace, not an opinion. Every report shows the numbered call sequence with the conflicting write marked.
- Do not fix inside the audit. Report, then hand off — an audit that starts editing loses its own coverage.
Growth path
This skill is intentionally single-file. If the defect-pattern list (Step 2)
grows past six patterns, or this file passes ~250-300 lines, split framework-
specific tracing detail (e.g. a Zustand/Redux-specific vs. a signals-specific
lens) into references/, following the domain+lens pattern other skills in
this ecosystem use — do not let a single growing file replace that split.
04 同仓库更多 Skills
video-style-edho-ferdian
Measure a reference video's look and pacing (colour grade as a 3D LUT, cut rhythm, overlays) and reproduce it: grade, cut and composite new footage, verify numerically, hand off an NLE timeline. Use for "tiru look video ini", "samakan grading", "bikin video dengan gaya referensi", "LUT dari referensi", "cut rhythm".
test-authoring-edho-ferdian
Write unit and component tests well: React Testing Library, pytest, Go and Vue, plus regression-test patterns. Use for "tulis test untuk component ini", "bagaimana test hook ini", "test yang bagus untuk fitur X", "tulis test pytest/Go/Vue", or dev-kickoff's TEST stage.
system-design-edho-ferdian
Mid-project architecture decisions: ADRs, trade-off analysis, non-functional requirements, scaling tiers and blast radius. Use for "desain arsitektur", "keputusan teknis besar", "bikin ADR", "trade-off antara X dan Y", "microservices atau monolith", "apa yang rusak kalau X diubah", or a capacity question.
spec-mining-edho-ferdian
Extract behavioural specs from a codebase with none: WHEN/THEN requirements and invariants, each anchored to the code that enforces it, unclear behaviour recorded as uncertainty. Use for "ekstrak spec", "buat spec dari kode", "reverse-engineer the spec", "repo ini tidak ada dokumentasi", or a missing BEHAVIOR_SPEC in dev-kickoff.
skill-authoring-edho-ferdian
Create and govern this ecosystem's skills: search before building, a quality bar, compliance checks, promoting repeated principles to rules, and packaging dist/*.skill. Use for "bikin skill baru", "ada skill buat X gak", "fork skill ini", "skill gue kepake gak sih", "package skill ini", "buatkan .skill-nya".
skill-audit-edho-ferdian
Audit this repo's skills/ for staleness, redundancy, broken cross-references and weak descriptions (not the ~/.claude setup: that is config-hygiene). Use for "audit skill saya", "cek skill yang sudah dibuat", "ada yang redundan gak", "skill mana yang basi".