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/johanthoren/jeff/agents-mdgit clone --depth 1 https://github.com/johanthoren/jeffWhat 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 | $0.01875 | $0.01875 |
| Opus 5 | $0.00937 | $0.00937 |
| Sonnet 5 | $0.00375 | $0.00375 |
| Haiku 4.5 | $0.00187 | $0.00187 |
Grade A, and why
jeff AGENTS.md 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 — 77 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md: jeff
jeff is a model-native quality control plane, distributed as Claude Code, Codex, and Cursor plugins and a Pi package. Grok Build consumes the Claude Code-compatible plugin surface directly. The method is the product, not a runtime: a thin orchestrator drives atomic tasks through fresh specialist contexts, with enforced separation, durable evidence, and deterministic gates. It is a cooperative workflow protocol for one trusted Chef and friendly frontier agents, not a security sandbox; host tool isolation is not a cross-host invariant.
Claude Code, Codex, Cursor, Grok Build, and Pi are first-class hosts over the same method and checked core; Oh My Pi installs the Pi package and uses its dispatch bridge. A Grok Bot loads Jeff through the shipped Cursor plugin. Grok Build is the coding client, while Grok is also a model family alongside Claude and GPT. Host mechanics differ; the method semantics stay shared.
Jeff 6.0 adds Graph Engineering at the work layer: fake-edge decomposition, task-DAG ready sets and atomic claims, isolated parallel lanes, facts-only context packets, write-ahead journaling, typed targeted repair, evidence-scaled convergence, and a versioned read-only projection. The implementation record and adjacent-system survey live in the 6.0 Graph Engineering slate.
- Design spec:
docs/specs/jeff-design.md - State schema:
skills/cook/reference/jeff-state-schema.md - Operational procedure:
skills/cook/SKILL.md - Voice/persona canon:
docs/brand.md
Before changing the system itself (the method, skills, agents, validator, dispatch policy), read docs/maintaining-jeff.md: the maintenance and model-drift stance. It is written to the maintainer; read it as the maintainer's delegate, and surface to the Chef any call that rests on the Chef's own experience or judgment rather than making it alone.
Model expectation
jeff is built for a frontier-tier model: Claude Opus 5, GPT-5.6 Sol, Grok 4.5, or a successor of that class. That is a design target, not a gate. The harness assumes that tier's judgment and carries no scaffolding to prop up less.
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 · 77 lines · 1,875 tokens per session scan A 550a7ea3b796
jeff AGENTS.md is an instructions file published in the GitHub repository johanthoren/jeff (4 stars, last pushed 5d ago), licensed Apache-2.0. It adds 1,875 tokens to every session, about $0.0094 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 instructions, from other repositories
ii-agent AGENTS.md
Instructions for Intelligent-Internet/ii-agent, covering ii-agent contributor guide, quick start, repository map, mandatory rules and architecture.
intelligent-terminal rust.instructions.md
Concise Rust coding conventions for this repository.
Browser4 CLAUDE.md
Instructions for platonai/Browser4, covering browser4 — project context for claude, architecture, key dispatch chain (cli → browser), batch commands and e2e test structure.
nimbalyst AGENTS.md
AGENTS.md instructions for nimbalyst/nimbalyst: At the beginning of every AI session, read the root CLAUDE.md and any applicable package-level CLAUDE.md before taking actions.
sortie copilot-instructions.md
Instructions for sortie-ai/sortie, covering sortie coding & review standards, 1. layered imports (downward only; violation is critical), 2. concurrency safety, 3. workspace path safety (critical: security boundary) and 4. persistence (sqlite).
GENesis-AGI AGENTS.md
AGENTS.md instructions for WingedGuardian/GENesis-AGI, covering agent instructions, code review mandate (adversarial), gitnexus — code intelligence (advisory), genesis capability surface and skills.