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 skills/madappgang/magus/agent-modelnpx skills add MadAppGang/magus --skill agent-modelgit clone --depth 1 https://github.com/MadAppGang/magusWrote 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/madappgang/magus/agent-model)<a href="https://agentmods.dev/skills/madappgang/magus/agent-model"><img src="https://agentmods.dev/badge/skills/madappgang/magus/agent-model.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 | $0.00044 | $0.01081 |
| Opus 5 | $0.00022 | $0.00541 |
| Sonnet 5 | $0.00009 | $0.00216 |
| Haiku 4.5 | $0.00004 | $0.00108 |
Grade C, and why
agent-model scanned grade C with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
`~/.claude/settings.json` → `<project>/.claude/settings.json` → Reads MCP configurationmediumAgent snooping
mcp.json carries server URLs and auth tokens; reading it lets a mod discover and abuse other integrations.
server: add it to the plugin's `.mcp.json` env block (e.g. `"ZHIPU_API_KEY": "${ZHIPU_API_KEY}"`) How it starts
The opening of the file, as written. The whole thing — 79 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Choosing the browser-use agent brain
The retry_with_browser_use_agent tool and cloud sessions run an autonomous agent loop that
needs its own LLM to think with (it makes its own API calls — it does not borrow this Claude
session). This skill decides which LLM, per task. The Python MCP server is deliberately dumb: it
reads config and accepts an override. The judgment lives here.
Selection precedence (how the server resolves the brain)
- Session override — set at runtime by the
browser_set_agent_modeltool (this turn only). - Configured — the
"browser-use"."agentModel"block merged from settings files:~/.claude/settings.json→<project>/.claude/settings.json→<project>/.claude/settings.local.json(local wins). - Legacy — if
BROWSER_USE_API_KEYis set and nothing above, Browser Use'sbu-latest. - Default —
anthropic/LATEST_SONNET_MODEL(usesANTHROPIC_API_KEY, already available).
Providers
| provider id | use for | key (env var NAME) | base_url |
|---|---|---|---|
anthropic |
Claude models — strong vision, best default | ANTHROPIC_API_KEY |
not needed |
openai |
GPT models | OPENAI_API_KEY |
not needed |
openai_compatible |
Kimi, GLM, Moonshot, Zhipu, or any OpenAI-compatible endpoint | your choice | required |
browser_use |
Browser Use's hosted BU3 models (bu-latest, bu-2-0) |
BROWSER_USE_API_KEY |
not needed |
Current Anthropic vision models: LATEST_SONNET_MODEL (default; best multi-step + vision),
LATEST_OPUS_MODEL (hardest reasoning), LATEST_HAIKU_MODEL (cheapest/fastest). Confirm live ids
against the claude-api skill before quoting one — do not rely on memory.
Picking a model for the task
- Simple scrape / read one page → cheap+fast is fine:
LATEST_HAIKU_MODEL, or a cheapopenai_compatiblemodel. - Multi-step form fill, navigation, recovering from popups →
LATEST_SONNET_MODEL(default). - Login / auth flows, ambiguous pages, hard reasoning →
LATEST_SONNET_MODELorLATEST_OPUS_MODEL. - CAPTCHA / stealth / anti-bot → a cloud session (
browser_start_cloud_session) withbrowser_use/bu-latest, which is tuned for it. - If the user names a model or provider (Kimi, GLM, "the cheap one", "smartest"), honor it.
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.
- 6d ago First seen · 79 lines · 44 tokens per session scan C 0e9e149b7085
agent-model is a skill published in the GitHub repository MadAppGang/magus (9 stars, last pushed 2d ago), licensed MIT. It adds 44 tokens to every session and 1,081 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it C with 2 findings (reads agent configuration directories, reads mcp configuration). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
v8-jit
V8 JIT optimization patterns for writing high-performance JavaScript in Next.js server internals. Use when writing or reviewing hot-path code in app-render, stream-utils, routing, caching, or any per-request code path. Covers hidden classes / shapes, monomorphic call sites, inline caches, megamorphic deopt, closure…
use-agent-browser-for-airi
Test AIRI display-model imports with agent-browser across stage-tamagotchi Electron, stage-web, and stage-pocket mobile web layouts. Use when uploading and verifying contributor-supplied Live2D ZIP, VRM, or MMD ZIP/PMX/PMD files through AIRI's model selector, including onboarding bypass, format-specific import…
opencli-sitemap-author
Use when creating or maintaining OpenCLI site sitemaps: agent-facing navigation, page-state, action, workflow, API-reference, pitfall, and fallback knowledge for a website. Use after browser exploration discovers durable site context, when a sitemap is stale, or when promoting local site knowledge into the repo.
debug-optimize-lcp
Guides debugging and optimizing Largest Contentful Paint (LCP) using Chrome DevTools MCP tools. Use this skill whenever the user asks about LCP performance, slow page loads, Core Web Vitals optimization, or wants to understand why their page's main content takes too long to appear. Also use when the user mentions…
webapp-testing
Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.
interactive-login
How to complete browser/interactive logins (aws / gh / glab / gcloud). The platform backgrounds the login poller so it survives the human's browser round-trip — and when that does NOT work.