Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add tmj-90/gaffer --skill performance-reviewgit clone --depth 1 https://github.com/tmj-90/gafferWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/tmj-90/gaffer/performance-review)<a href="https://agentmods.dev/skills/tmj-90/gaffer/performance-review"><img src="https://agentmods.dev/badge/skills/tmj-90/gaffer/performance-review/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/tmj-90/gaffer/performance-review"><img src="https://agentmods.dev/badge/skills/tmj-90/gaffer/performance-review.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00093 | $0.01226 |
| Opus 5.5 | $0.00037 | $0.00490 |
| Sonnet 5.5 | $0.00019 | $0.00245 |
| Haiku 4.5 | $0.00009 | $0.00123 |
Grade A, and why
performance-review scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 16d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 80 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Performance review lens
Performance regressions ship because each looks reasonable alone: one more query in a
loop, one more sequential await, one more import. For every new hot-path line ask: how
many times does this run, what does each run cost, and what bounds it? A finding is a
concrete defect only when it is unbounded (grows with input an outside party
controls), stalls an async server's request path (a synchronous I/O call, an outbound
call with no timeout), or measurably breaks a budget the repo or an AC states;
everything else is a note. Write the arithmetic into every finding.
Steps
- Find the hot paths in the diff: request and job handlers, loops over collections
whose size is not fixed, render functions, anything called per item or per request.
Skip one-off setup unless it runs at startup of a hot service. Call
search_lorefor known hot spots, budgets and the expected data sizes. - Count and cost each path against the checklist: "this query runs once per order on the page; a page holds up to 200 orders; that is 201 queries per request". Use the largest size the code allows, not the size in the test fixture.
- Check the bound. Is the collection capped (pagination,
LIMIT, a max upload size, a validated array length)? If a user, tenant or upstream can make it arbitrarily large, the work and memory are unbounded. - Check performance claims. A ticket whose AC promises a speedup or a latency target
needs before/after numbers under stated conditions with repeated runs (the
performance-profilingskill); without them that AC is unevidenced. - Rate each finding. Blocking: unbounded work or memory reachable from input; an N+1
on a list path whose size is unbounded; a blocking call on the
request path of an async server; a missing timeout on an outbound call in a request
path; a budget the repo or an AC enforces being broken. Note: a bounded regression
that breaks no stated budget, or an optimisation opportunity. Only blocking findings
justify
RECOMMEND CHANGES. - Record each finding with
record_ac_evidence(evidence_type: manual_note, with the arithmetic and the fix), then let thereview-ticketverdict carry the result.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 16d ago Changed · +10 lines f94f5a29c007
- 19d ago First seen · 70 lines · 93 tokens per session scan A 55df743d064a
performance-review is a skill published in the GitHub repository tmj-90/gaffer (2 stars, last pushed 17d ago), licensed Apache-2.0. It adds 93 tokens to every session and 1,226 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-28.
Other skills, from other repositories
test-audit
Audit test suites for T1-T4 violations using AST analysis, mock detection, and multi-stage synthesis. Invoke when user asks to audit tests, check test quality, find mock violations, review test effectiveness, or inspect test suites for over-mocking. Triggers automatic rewrites when quality gates fail.
code-review
Comprehensive code review with distinct aspect based sections. Use when reviewing code, checking for security issues, finding type safety problems, auditing code quality, or when user asks to review code, PRs or changes. Three-phase workflow runs static tools, LLM judgment, and writes diagnostic log.
spec-drift-check
Audits a WP brief for drift against current code — extracts claims, verifies each, emits PROCEED/STOP verdict. Use when starting a new WP, before consuming a spec as binding, or when a doc references paths/lines/functions.
besimple-broccoli-blind
Non-interactive wrapper: plan-sketch -> auto-pick recommended options -> plan-write -> plan-critique-loop -> implement-from-plan -> claude-simplify-wrapper -> dedup -> code-review-loop. No PR creation and no Linear comments.
code-review-loop
Iterative review+fix loop for BASESHA..HEAD: generate findings, apply accepted fixes, run checks, commit, and re-review up to 3 iterations or until clean.
dedup
Dedupe-only pass for BASESHA..HEAD: remove duplicate code introduced by the diff or reuse existing shared utils; applies changes + commits.