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/memgrafter/flatmachines/agents-mdgit clone --depth 1 https://github.com/memgrafter/flatmachinesWhat 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 | $0.02185 | $0.02185 |
| Opus 5 | $0.01092 | $0.01092 |
| Sonnet 5 | $0.00437 | $0.00437 |
| Haiku 4.5 | $0.00218 | $0.00218 |
Grade A, and why
flatmachines 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 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.
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 — 217 lines — stays where its author put it; the contents beside it link to each section on GitHub.
FlatAgents + FlatMachines Reference
Target: <1000 tokens. LLM-optimized. See
flatagent.d.ts,flatmachine.d.ts,profile.d.ts,prompt.d.tsfor schemas.Versioning: All specs and SDKs use lockstep versioning.
Concepts
FlatAgent: Single LLM call. Model + prompts + output schema. No orchestration. FlatMachine: State machine orchestrating agents. States, transitions, conditions, loops, error handling.
| Need | Use |
|---|---|
| Single LLM call | FlatAgent |
| Multi-step/branching/retry/errors | FlatMachine |
| Parallel execution | machine: [a, b, c] |
| Dynamic parallelism | foreach |
| Background tasks | launch |
Model Profiles
# profiles.yml — agents reference by name
spec: flatprofile
spec_version: "4.2.1"
data:
model_profiles:
fast: { provider: cerebras, name: zai-glm-4.6, temperature: 0.6 }
smart: { provider: anthropic, name: claude-3-opus-20240229 }
default: fast # Fallback
# override: smart # Force all
Agent model field: "fast" | { profile: "fast", temperature: 0.9 } | { provider: x, name: y }
Resolution: default → profile → overrides → override
OAuth Backends (Codex / Copilot)
backend: codexandbackend: copilotare explicit-only (never auto-detected).- Backend selection precedence remains: constructor
backend→ resolvedmodel.backend→ auto-detect (litellm/aisuite only). oauthsettings are read from resolved model config; works identically whether model came from inline agent config or profile.- Auth file precedence:
- Codex:
oauth.auth_file→ legacycodex_auth_file→ legacyauth.auth_file→FLATAGENTS_CODEX_AUTH_FILE→~/.pi/agent/auth.json - Copilot:
oauth.auth_file→copilot_auth_file→ legacyauth.auth_file→FLATAGENTS_COPILOT_AUTH_FILE→~/.agents/flatmachines/auth.json
- Codex:
- Token handling: pre-request refresh on expiry; if refresh fails, re-read auth store once for cross-process refresh; fallback refresh+retry on
401/403. - Retries on
429/500/502/503/504with exponential backoff (no jitter).
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 · 217 lines · 2,185 tokens per session scan A 1566e79fca1a
flatmachines AGENTS.md is an instructions file published in the GitHub repository memgrafter/flatmachines (37 stars, last pushed 3mo ago), licensed Apache-2.0. It adds 2,185 tokens to every session, about $0.0109 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 instructions, from other repositories
agentcontrolplane CLAUDE.md
Claude Code instructions for humanlayer/agentcontrolplane, covering agent control plane (acp) - ai assistant instructions, 🚨 the 1500-line minimum read rule - this is not optional, 📋 your 20-point todo list - you need this structure, current todo list (you must maintain 20+ items) and 🚨 mandatory persona selection.
graphiti CLAUDE.md
Instructions for getzep/graphiti, covering claude.md, project overview, development commands, main development commands (run from project root) and install dependencies.
stagehand AGENTS.md
Instructions for browserbase/stagehand: Only a human may request changes to this file. Keep additions rare and limited to durable, repository-wide rules. -->.
agent-framework copilot-instructions.md
Instructions for microsoft/agent-framework, covering github copilot instructions, repository structure and architectural decision records (adrs).
agent-framework python.instructions.md
Instructions for microsoft/agent-framework: See AGENTS.md for project structure and package documentation. Detailed conventions are in the agent skills under .github/skills/.
graphiti AGENTS.md
Instructions for getzep/graphiti, covering repository guidelines, project structure & module organization, build, test, and development commands, coding style & naming conventions and testing guidelines.