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.
git clone --depth 1 https://github.com/JoaquinRuiz/SpecJudgeWrote 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/commands/joaquinruiz/specjudge/speckit.specjudge.recommend)<a href="https://agentmods.dev/commands/joaquinruiz/specjudge/speckit.specjudge.recommend"><img src="https://agentmods.dev/badge/commands/joaquinruiz/specjudge/speckit.specjudge.recommend/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/commands/joaquinruiz/specjudge/speckit.specjudge.recommend"><img src="https://agentmods.dev/badge/commands/joaquinruiz/specjudge/speckit.specjudge.recommend.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 | $0.00019 | $0.00941 |
| Opus 5 | $0.00010 | $0.00470 |
| Sonnet 5 | $0.00004 | $0.00188 |
| Haiku 4.5 | $0.00002 | $0.00094 |
Grade A, and why
speckit.specjudge.recommend 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 11d 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 — 88 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Recommend a model for this work
The tasks exist and nothing has been implemented yet. That is the one moment where choosing a model is both informed and free — the artifacts are written, and no tokens have been spent. SpecJudge reads what is already in the project and says which model fits it, citing the fragment behind each level.
This command reports; it does not decide and does not write. Nothing in the project changes, and the user is free to ignore the answer.
User Input
$ARGUMENTS
Treat any argument as an alternative project path. With no argument, use the current project root.
Prerequisites
specjudgeonPATH(uv tool install specjudge, orpip install specjudge).- Ollama running with a local model, and a judge already chosen. SpecJudge remembers the
choice, so this is a one-off:
specjudge --set-judge.
If either is missing, say so in one line and stop. Do not install anything, and do not block the spec-kit flow — the user came here to implement, not to configure a recommender.
Execution
-
Run, from the project root:
specjudge . --json --execution-model singlesingleis deliberate: it ranks on the hardest part of the work, which is the correct reading when nobody is present to say whether they can switch model per task. -
Handle the exit code before reading anything:
Exit Meaning What to do 0 A recommendation was produced Report it (step 3) 2 Not enough project information Say the tasks are too thin to judge, and continue 3 Judge unavailable (Ollama missing, no model, not configured) Say what is missing, in one line, and continue 4 Model catalog missing or empty Report it as a SpecJudge installation problem, and continue Any non-zero exit ends this command without stopping whatever the user was doing. A recommendation nobody asked for must never be the reason an implementation did not start.
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.
- 11d ago First seen · 88 lines · 19 tokens per session scan A e73dfb9ff771
speckit.specjudge.recommend is a command published in the GitHub repository JoaquinRuiz/SpecJudge (47 stars, last pushed 7d ago), licensed MIT. It adds 19 tokens to every session and 941 once invoked, about $0.0001 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 2026-08-30.
Other commands, from other repositories
compress
Automatically compress CLAUDE.md or other memory files with protected-span safety. One-command token savings.
crystallize
Transform verbose SOPs into high-performance agentic prompts via principled compression. Use when user says "/crystallize", "compress this prompt", "make this more agentic".
prompt
Transform the user message into a professional, precise Cursor-ready prompt. Never execute the requested task. Output only the optimized prompt package.
t00-ai-dev
A set of instructions for building applications that use artificial intelligence, such as chatbots, document search, or text-generation tools. It covers Claude, retrieval-augmented generation (RAG), embeddings, and vector databases, which help find relevant documents for an AI answer.
simplicio
Run the next user task through the simplicio-prompt Tuple-Space + Yool safe-speed runtime.
distill
Convert a vague or under-specified request into a clear, efficient AI-ready instruction.