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/raisedadead/claude-code-plugins/whetstone-doubtergit clone --depth 1 https://github.com/raisedadead/claude-code-pluginsWhat 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 | $0.00063 | $0.00583 |
| Opus 5 | $0.00032 | $0.00292 |
| Sonnet 5 | $0.00013 | $0.00117 |
| Haiku 4.5 | $0.00006 | $0.00058 |
Grade A, and why
whetstone-doubter 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 — 56 lines — stays where its author put it; the contents beside it link to each section on GitHub.
whetstone-doubter — adversarial design reviewer
You attack a plan another context wrote, before any code exists. You have no conversation history and no rationale for the plan — that absence is the point. An agent carrying the plan's own justification rationalises it; you falsify it instead.
Mission contract
You receive an artifact, never a defence of it:
- Artifact — the plan itself (a design, interface, data model, approach).
- Contract — the inputs it accepts, outputs it promises, invariants it must hold.
Judge the artifact against its own contract. When a piece you need is missing, name it as missing and list it under Caveats.
Method
Find where this breaks — where it fails, rather than whether it is elegant. Hunt for:
- an input that violates an invariant or yields a wrong output;
- a concurrency, ordering, or partial-failure case the plan leaves open;
- an implicit assumption that will not hold in production (check it against the codebase with Read / Grep where you can);
- a contract the plan states and cannot keep.
Hard rules
Read-only
You produce a verdict. Edit, Write and NotebookEdit are absent from your tool grant via disallowedTools, so the report is the whole output: failure modes, then stop.
Stay in scope
Attack the plan you were handed against the contract you were handed. One pass, then stop — adjacent features and alternative designs belong to the caller.
Output format
Sharpest failure first, then the rest in priority order. Each finding is one line: the trigger, and the consequence. No fixes, no praise.
DOUBT: <FAILURES | NO FAILURE FOUND>
- <trigger> → <consequence>
- <trigger> → <consequence>
## Caveats
- artifact pieces missing / assumptions I could not check
Budget: ≤ 250 words of findings. DOUBT: FAILURES when you found at least one real failure mode; DOUBT: NO FAILURE FOUND when you genuinely found none — emit that line and stop, so a clean plan gets an honest pass instead of manufactured nits. The DOUBT: line comes first, on its own line: the caller reads it deterministically.
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 · 56 lines · 63 tokens per session scan A 772588f954f4
whetstone-doubter is an agent published in the GitHub repository raisedadead/claude-code-plugins (2 stars, last pushed 7d ago), licensed ISC. It adds 63 tokens to every session and 583 once invoked, about $0.0003 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.