The broad latency pass — wasted renders, main-thread stalls, startup waterfalls — ranked by what the user actually feels on common paths.
Free & open · no signup · read-only — it ends by asking · nothing leaves your machine
You are working inside this repo. Mission: find where this product wastes time and bytes, and rank the wins by latency the user actually feels on common paths.
This is the broad pass across every layer. When the database is the suspect, go deep with 87; when the payload is, 88; when caching is, 140.
Read-only pass: build, measure, profile — but change nothing. Your only write is the report file.
Phase 2 sweeps the codebase through every one of these, citing file and line for each finding.
bundle size, missing code-splitting, unoptimized images/fonts, oversized JSON responses
unnecessary re-renders, unmemoized expensive work, layout thrash, long lists without virtualization
N+1 patterns, over-fetching, queries in loops, missing indexes implied by access patterns
repeated identical fetches/computations with no memo or cache layer
heavy synchronous work where users are waiting
cold-start waterfalls, sequential awaits that could be parallel, eager work that could be lazy
unbounded caches, retained references, growing arrays
One structured report at the repo root — or in reports/, if you keep one — the same shape every time, ready for a teammate — or the optional Studio — to act on.
Every Goal Prompt follows the same four steps, so results are consistent and repeatable — no matter which one you run.
Build the app.
Cite location and estimated cost for every finding.
Rank by user-felt impact on common paths — not micro-benchmarks.
Create PERF.md at repo root.
Copy it and paste it into your agent inside the repo you want checked.
Install the goal plugin once — two commands — then just type /goal:performance-audit.
/plugin marketplace add GhostlyGawd/goal-prompts/plugin install goal@goal-promptsOr install only this Goal Prompt as /goal-performance-audit:
curl -fsSL https://goal-prompts.vercel.app/install | BRIEF=04 shLet an agent fetch it mid-conversation, or pull the raw Goal Prompt by URL.
https://goal-prompts.vercel.app/raw/04.mdNothing hidden — this is the whole Goal Prompt, verbatim. Read it in a minute, edit it, or copy it as-is.
# Goal: Performance Audit You are working inside this repo. Mission: find where this product wastes time and bytes, and rank the wins by latency the user actually feels on common paths. This is the broad pass across every layer. When the database is the suspect, go deep with 87; when the payload is, 88; when caching is, 140. Read-only pass: build, measure, profile — but change nothing. Your only write is the report file. ## Phase 1 — Measure what's measurable - Build the app; record bundle sizes and build time. - Time the test suite and any measurable startup path. - Identify the 3 heaviest pages/endpoints by code inspection: most data fetched, most rendered, most computed. ## Phase 2 — Audit through 7 lenses Cite location and estimated cost for every finding. 1. **Payload** — bundle size, missing code-splitting, unoptimized images/fonts, oversized JSON responses 2. **Render waste** — unnecessary re-renders, unmemoized expensive work, layout thrash, long lists without virtualization 3. **Query cost** — N+1 patterns, over-fetching, queries in loops, missing indexes implied by access patterns 4. **Missing caches** — repeated identical fetches/computations with no memo or cache layer 5. **Main-thread blockers** — heavy synchronous work where users are waiting 6. **Startup cost** — cold-start waterfalls, sequential awaits that could be parallel, eager work that could be lazy 7. **Memory growth** — unbounded caches, retained references, growing arrays ## Phase 3 — Curate - Rank by user-felt impact on common paths — not micro-benchmarks - For each finding, estimate the gain: order of magnitude is fine (ms saved, kb removed) - Kill anything that trades meaningful complexity for negligible gain ## Phase 4 — Report Create `PERF.md` at repo root: 1. **Current numbers** — everything measured in Phase 1 2. **Findings** — each: Name · Location · Cost today (estimate) · Fix · Expected gain · Effort S/M/L 3. **Quick wins vs deep work** — the split, with the top quick win spelled out 4. **Measurement plan** — how to verify each gain after fixing Start the report with today's date. If `PERF.md` already exists from a previous run, read it first and lead with what changed since. ## Rules - A perf claim needs a number or a mechanism — no vibes - Optimize the common path before the edge case - No runtime hot paths in this repo? Say so in a one-paragraph null report and stop — a null result is a valid finding. - If a `reports/` directory exists at the repo root, write the report there instead of the root. - Before asking, present the top findings as a ranked list in plain words - Report only — end by asking which wins to take
reports/ directory exists at the repo root, write the report there instead of the root.Curated neighbors — briefs that answer the adjacent question, worth running in the same session.
The database access patterns that fall over as data and traffic grow — N+1s, missing indexes, unbounded reads, and lock contention.
What the browser actually downloads — oversized bundles, unsplit code, heavy dependencies, and unoptimized assets that slow first paint.
Every cache from CDN headers to that module-level dict — keys, TTLs, who invalidates what — the stale read, the stampede, and the key that leaks between users.
Simulate 10x users, data, and traffic on paper — find what breaks first, at what threshold, and the cheapest mitigation.
Where the seconds go. Decomposes p50 and p95 of the runs users feel into stages — queue, retrieval, first token, tools — and names the stage worth attacking.
The database access patterns that fall over as data and traffic grow — N+1s, missing indexes, unbounded reads, and lock contention.
What the browser actually downloads — oversized bundles, unsplit code, heavy dependencies, and unoptimized assets that slow first paint.
Runs inside these playbooks — curated sequences you can launch with one paste: