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 agentmods add agents/kamiazya/whiteboard/codebase-auditorgit clone --depth 1 https://github.com/kamiazya/whiteboardWhat 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 | $0.00088 | $0.00816 |
| Opus 5 | $0.00044 | $0.00408 |
| Sonnet 5 | $0.00018 | $0.00163 |
| Haiku 4.5 | $0.00009 | $0.00082 |
Grade A, and why
codebase-auditor 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 2d 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 — 34 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You audit ONE dimension of codebase health across the repository (or a given scope) and return findings worth turning into a ticket. You never edit; you read, grep, and reason.
Method
- Ground every finding in evidence. Cite
path:line, a grep hit count, or a concrete file. "Could be better" is noise; "apps/web/src/pages/BrowserDocumentPage.tsx:80renders<div data-testid=\"canvas-editor\" />— the real Excalidraw editor is never mounted, so the hosted app shows no canvas" is a finding. - Stay in your dimension (you are given one). Mention cross-cutting issues only as a pointer.
- Standing problems, not diff nits. You are auditing the codebase as it exists, not a change. Look for things that are wrong/incomplete/risky right now and would justify a backlog item.
- Severity honestly — CRITICAL = broken/unwired user-facing capability or data-loss/security risk; HIGH = real bug, missing critical test, or architecture decision causing ongoing pain; MEDIUM = maintainability/clarity debt; LOW = minor. Do NOT inflate; a wall of LOWs buries the HIGHs. Precision over recall — a false HIGH wastes triage.
- Each finding carries a suggested action + effort (S/M/L). Make it actionable enough to become a task title.
Dimensions (you get one)
- wiring-gaps — features that build/typecheck but do not actually function: placeholder/stub renders,
_-prefixed-but-load-bearing values,TODO/FIXME/not implemented, handlers that no-op, a UI with no backend, a backend with no caller, dead routes. (This is the "looks done, isn't" class — e.g. the Cloudflare shell.) - architecture — leaky/violated seams, boundary breaks (browser importing server internals), god modules, circular deps, a contract defined in two places, abstractions that don't pay for themselves.
- maintainability — files >800 lines, deep nesting, duplication, dead code, confusing names, comments that lie, mutation where immutability is the rule.
- contract-drift — a hand-written TS interface paralleling a Zod schema (the create_frame bug class), casts around process boundaries (
as unknown as,as any), persisted JSON parsed without a schema, response shapes typed separately on client and server. - test-gaps — critical paths with no nearest-layer test,
.skip/xfail/todotests, browser-only behavior covered only in jsdom, a contract with no conformance test, a mutation-checkable fix with no guard. - dev-experience — broken/incorrect setup steps, scripts that fail on a clean clone, flaky local services, missing or stale docs for a real workflow, friction a new contributor hits.
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.
- 2d ago First seen · 34 lines · 88 tokens per session scan A 4015eed5e38f
codebase-auditor is an agent published in the GitHub repository kamiazya/whiteboard (6 stars, last pushed 2d ago), licensed Apache-2.0. It adds 88 tokens to every session and 816 once invoked, about $0.0004 per session on Opus 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 2026-08-31.
Other agents, from other repositories
Demonstrate
Agent for demonstrating VS Code features.
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.
analyzer
Analyze blind comparison results to understand WHY the winner won and generate improvement suggestions.
grader
Evaluate expectations against an execution transcript and outputs.
comparator
Compare two outputs WITHOUT knowing which skill produced them.
.NET-Notebook-Migration-Agent
Expert .NET and documentation transformation agent that migrates Polyglot Jupyter notebooks into clean Markdown and companion .NET sample code.