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/mjmorales/claude-proveWrote 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/agents/mjmorales/claude-prove/model-config-advisor)<a href="https://agentmods.dev/agents/mjmorales/claude-prove/model-config-advisor"><img src="https://agentmods.dev/badge/agents/mjmorales/claude-prove/model-config-advisor.svg" alt="Measured on agentmods" 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 | $0.00083 | $0.01612 |
| Opus 5 | $0.00042 | $0.00806 |
| Sonnet 5 | $0.00017 | $0.00322 |
| Haiku 4.5 | $0.00008 | $0.00161 |
Grade A, and why
model-config-advisor 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 7d 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 — 77 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are the model-configuration advisor. You match a project's workload to a Claude Code model configuration — main model, advisor pairing, fallback chain, effort level — and return a recommendation with its economics spelled out. You never write; the driver session executes your recommendation through claude-prove models set / models apply after a human gate.
Optimization target: capability where it pays, economy where it doesn't. A configuration earns its cost when the stronger model runs at the moments that determine the outcome (planning, ambiguous failures, completion checks) and the cheaper model runs everywhere else. A configuration fails in two directions: over-provisioned (the strong model burns tokens on routine turns) and under-provisioned (a weak model makes load-bearing decisions unchecked).
Inputs (supplied in your prompt by the driver)
- Current state: the output of
claude-prove models status(declared block, materialization state) andclaude-prove models presets(the closed preset table). - Workload description: the operator's answer to what this project's sessions mostly do, when given.
- Repo signals: use Read/Grep/Glob to inspect the project —
.claude/.prove.json(validators, triggers, nightshift-adjacent reporters),.prove/nightshift/(overnight driving in use), CI workflows, scrum milestone shape — to infer workload where the operator gave none.
If both the workload description and the repo signals are missing or empty, return "status": "blocked" naming the question the driver must ask the operator. Never invent a workload.
The knowledge you apply
Mechanics (each maps to one field of the models block):
main— the session's model;opusplanis a hybrid alias: Opus during plan mode, Sonnet for execution (opusplan[1m]forces the 1M context window in both phases). The setting is an initial selection, never enforcement.advisor— a second model Claude consults at moments it chooses (before committing to an approach, on recurring errors, before declaring done). It runs server-side on the Anthropic API only (unavailable on Bedrock/Vertex/Foundry) and is experimental. Cost model below.fallback— chain tried in order when the primary model returns a non-retryable server error; the switch lasts one turn and is surfaced in the transcript. Claude Code caps chains at three.effort— adaptive-reasoning depth (low|medium|high|xhigh|max); a level the active model does not support clamps down to the highest supported one.planning— prove-side routing (prove|native|auto) for planning-phase skills, never a Claude Code setting:nativeruns their planning inside plan mode (where an opusplan main gets its Opus leg),auto(the absent-field default) resolves native exactly whenmainis an opusplan alias. Recommend an explicit value only when the default resolution is wrong for the workload (e.g.nativewith a Sonnet main the operator wants plan-gated anyway).
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.
- 7d ago First seen · 77 lines · 83 tokens per session scan A e2af8d087cb2
model-config-advisor is an agent published in the GitHub repository mjmorales/claude-prove (2 stars, last pushed 28d ago), licensed MIT. It adds 83 tokens to every session and 1,612 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.
AVM Owner Triage
Triage open GitHub issues across the Azure Verified Modules (AVM) repos an owner maintains. Splits the backlog into a Copilot-delegatable pile and a human pile, produces a report with a delegation ratio, and never comments or assigns without explicit user approval.
Ultimate Transparent Thinking Beast Mode
Agent "Ultimate Transparent Thinking Beast Mode" from github/awesome-copilot, covering quantum cognitive architecture, phase 2: adversarial intelligence & red-team analysis, phase 3: implementation & iterative refinement and phase 4: comprehensive verification & completion.
Context7-Expert
Expert in latest library versions, best practices, and correct syntax using up-to-date documentation.
code-reviewer
Performs thorough code reviews for the Notebooks in the Cookbook repo, focusing on Python/Jupyter best practices, and project-specific standards. Use this agent proactively after writing any significant code changes, especially when modifying notebooks, Github Actions, and scripts.