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/atlyslabs/gideon/agents-mdgit clone --depth 1 https://github.com/atlyslabs/gideonWhat 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 | $0.04243 | $0.04243 |
| Opus 5 | $0.02122 | $0.02122 |
| Sonnet 5 | $0.00849 | $0.00849 |
| Haiku 4.5 | $0.00424 | $0.00424 |
Grade A, and why
gideon 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 — 387 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Setup and orientation guide for Gideon. Written for whoever (or whatever) is standing this up for the first time: what the features are, which ones need credentials, and exactly which credentials.
The short version: three environment variables get you a working instance. Everything else is optional, and every integration you leave blank simply does not appear as a tool. Nothing breaks, the tool just is not offered to the model.
1. Fastest path to a running instance
git clone https://github.com/atlyslabs/gideon.git
cd gideon
cp backend/.env.example backend/.env
Edit backend/.env and set three things:
| Variable | How to get it |
|---|---|
ANTHROPIC_API_KEY |
console.anthropic.com. Or use OPENAI_API_KEY / KIMI_API_KEY instead, see section 5. |
ENCRYPTION_KEY |
python -c "from cryptography.fernet import Fernet; print(Fernet.generate_key().decode())" |
SINGLE_USER_EMAIL |
Your email. Used as the local identity. |
Then:
docker compose up
Open http://localhost:3000. Postgres, Redis, the API and the frontend all come
up together, and the schema is applied on first boot from
backend/migrations/init.sql.
You are now logged in as one user with no password. That is AUTH_MODE=single_user,
the default, and it is the intended setup for personal and self-hosted use.
Do not expose an instance in this mode to a network you do not control,
because there is no authentication in front of it.
Without Docker
# backend (Python 3.13)
cd backend && poetry install && poetry run uvicorn main:app --reload --port 8080
# frontend (Node 20)
cd frontend && npm install && npm run dev
Postgres and Redis still have to be reachable at whatever backend/.env points at.
2. Features, and what each one costs you in setup
"Needs nothing" means it works on a fresh instance with only the three required variables above.
| Feature | What it does | What it needs |
|---|---|---|
| Chat | Streaming chat, extended thinking, tool use, file understanding | Needs nothing |
| Memory | Remembers durable facts about you, recalls them in later chats | Needs nothing. docker compose runs the Neo4j that powers the graph view; without it memory still works, backed by Postgres |
| Todos and delegation | Todos, plus delegating one to people over a Slack group DM with recurring nudges | Todos need nothing. Delegation needs Slack |
| Workflows | Natural-language automations on a schedule (every N minutes to weekly) | Needs nothing, but a workflow is only as useful as the integrations it can reach |
| Skills | Reusable instructions Gideon learns and applies | Needs nothing |
| Custom slash commands | Your own /<slug> prompts |
Needs nothing |
| Hub | Daily briefing: meetings, todos, suggestions | Needs nothing. Meetings need Google, weather needs WEATHER_API_KEY |
| Usage and cost | Per-model token and USD accounting | Needs nothing |
| In-chat React previews | Gideon writes a component, you see it rendered | Needs a separate compile service, see PREVIEW_BUILDER_URL in section 7. Returns 501 unset |
| Meeting prep | Auto-generated prep notes before a meeting | Google (Calendar) |
| Email auto-draft | Drafts replies in your voice | Google (Gmail) |
| Slack trigger | @-mention Gideon in Slack, it replies in thread | GIDEON_BOT_TOKEN |
| Email trigger | CC a mailbox to kick off a run | GIDEON_INBOX_EMAIL + Google |
| WhatsApp trigger | Message Gideon on WhatsApp | WHATSAPP_* (Meta Cloud API) |
| Analytics over your warehouse | Gideon writes SELECT-only SQL against ClickHouse | CLICKHOUSE_*, see section 6 |
| Surveys | Admin-built in-app surveys with cohort targeting | SURVEY_ADMINS (comma-separated emails) |
| Web search | Exa-backed research | EXA_API_KEY |
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 · 387 lines · 4,243 tokens per session scan A dd44f0c47c4a
gideon AGENTS.md is an instructions file published in the GitHub repository atlyslabs/gideon (7 stars, last pushed 20d ago), licensed Apache-2.0. It adds 4,243 tokens to every session, about $0.0212 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
outsystems-mcp CLAUDE.md
Instructions for OutSystems/outsystems-mcp, covering claude.md, supported harnesses, validate every change against every supported harness, skill docs must stay in lockstep across hosts and check before opening a pr.
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).