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 instructions/trueoriginlabs/vibatchium/agents-mdgit clone --depth 1 https://github.com/trueoriginlabs/vibatchiumWhat 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 | $0.09802 | $0.09802 |
| Opus 5 | $0.04901 | $0.04901 |
| Sonnet 5 | $0.01960 | $0.01960 |
| Haiku 4.5 | $0.00980 | $0.00980 |
Grade B, and why
vibatchium AGENTS.md scanned grade B with 1 finding 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 3d 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.
Instruction-override phrasingmediumPrompt injection
Text telling the model to disregard its earlier instructions or safety rules is the shape of a prompt injection, whoever wrote it.
$VB safety scan "ignore previous instructions and …" # test a string against the classifier Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.
How it starts
The opening of the file, as written. The whole thing — 591 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — vibatchium agent contract
If you're a coding agent (Codex, Cursor, Claude Code) and a user said "use vibatchium," read this. Saves ~15 min of environment-discovery friction.
First-time setup (for users)
pipx install 'git+https://github.com/trueoriginlabs/vibatchium#egg=vibatchium[all]' # core install drops the [all] for browse-only
patchright install chrome # optional preflight — the first launch auto-installs Chrome if missing
vb setup # wire vibatchium into Codex / Claude Code / Cursor (idempotent)
vb install # verify: prints core readiness + which optional lanes (fetch/vision/secrets/…) are available
After setup, any agent session in any cwd sees vibatchium as a registered MCP server. Restart agent sessions to pick up the registration.
vb fetchandvb search(the curl_cffi TLS-fingerprint HTTP lane) need the[fetch]extra. A core-only install can browse butvb fetchwill say which interpreter to add curl_cffi to. On a uv venv (no pip):uv pip install --python <venv>/bin/python curl_cffi. NB the extra is a property of the daemon's venv, not yours — whichever venv spawned the shared daemon decides whether the lane imports.
Staying current (read this if vb came from a git clone)
git pull updates the source. Three things downstream of it can keep serving the old world, and only one of them announces itself:
| What's stale | How to see it | Fix |
|---|---|---|
| The binary — a non-editable install copied the source, so a pull changes nothing | vb --version disagrees with git describe --tags (no warning otherwise) |
uv pip install -e '.[all]' |
| The daemon — long-lived, still executing the code it imported at boot | vb status → stale_code: true (the version compare can't see this: a pull doesn't move __version__) |
vb shutdown — the next vb call respawns |
The MCP cap set — frozen into your agent's config at first registration, so buckets added later (0.19.0's search) never appear |
vb setup reports the drift |
vb setup --force --caps lean,search, then restart the agent session |
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.
- 3d ago First seen · 591 lines · 9,802 tokens per session scan B c5b8e9ec0e4e
vibatchium AGENTS.md is an instructions file published in the GitHub repository trueoriginlabs/vibatchium (5 stars, last pushed 7d ago), licensed Apache-2.0. It adds 9,802 tokens to every session, about $0.0490 per session on Opus 5. A static security scan graded it B with 1 finding (instruction-override phrasing). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
camofox-browser copilot-instructions.md
Instructions for redf0x1/camofox-browser, covering camofox browser — github copilot instructions, what camofox is, core workflow (snapshot-first), most-used cli commands and lifecycle / navigation.
camofox-browser GEMINI.md
Instructions for redf0x1/camofox-browser, covering gemini.md — camofox quick rules, key rules and context.
mcp-camoufox AGENTS.md
Instructions for RobithYusuf/mcp-camoufox, covering mcp-camoufox — agent guide, product and repository map, hard constraints, working defaults and architecture invariants.
stealth-cli AGENTS.md
Instructions for Youhai020616/stealth-cli, covering stealth-cli, project structure, key architecture decisions, camoufox-js api (do not guess) and error handling.
stealth-chrome-devtools-mcp CLAUDE.md
Claude Code instructions for DevinoSolutions/stealth-chrome-devtools-mcp, covering claude.md — navigation map for agents, the four conventions (non-negotiable), navigation map (the tree as it is), top-level and embedded/ — the backend.
eyebrowse CLAUDE.md
Claude Code instructions for Evil-Bane/eyebrowse, covering claude.md — eyebrowse build notes, layout, run, pinned versions (lockstep matters) and key constraints (encoded in code — don't "simplify" away).