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/rux-eth/iris/claude-mdgit clone --depth 1 https://github.com/rux-eth/irisWhat 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 | $0.01373 | $0.01373 |
| Opus 5 | $0.00687 | $0.00687 |
| Sonnet 5 | $0.00275 | $0.00275 |
| Haiku 4.5 | $0.00137 | $0.00137 |
Grade A, and why
iris CLAUDE.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 — 86 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
What Is This Project
Iris — a self-hosted, chat-based personal life-management assistant for a single user (the owner),
built on OpenClaw and driven by Claude. (Named for the Greek messenger goddess; persona lives in the
workspace SOUL.md/IDENTITY.md, established PR-017.) The user talks to it over Telegram; it owns
their Google Calendar, manages tasks in Todoist, produces a priority-ranked daily plan,
surfaces reminders, and keeps competing demands (classes/assignments, work shifts, social plans,
health, errands) in balance. It both advises and acts — cautiously at first, with autonomy
expanding over time as trust is earned. Runs in Docker on the user's always-on Ubuntu desktop,
reachable only over Tailscale. A partner may be added later (architecture supports it).
Build & Test Commands
Stack is Node.js + TypeScript (OpenClaw + our custom MCP server). Commands will be established by PR-001/PR-007 (no code exists yet). Expected shape:
- Lint/typecheck/test: standard
npmscripts (defined when the MCP server lands). - Deploy + smoke test: see
docs/DEPLOYMENT.md. - Operate OpenClaw:
openclaw security audit,openclaw secrets audit --check,OPENCLAW_CACHE_TRACE=1for cache diagnostics.
Architecture
Hybrid "own the brain, rent the surfaces": an owned Markdown brain (GOALS.md/PRIORITIES.md) +
custom owned-core MCP server is the source of truth; Google Calendar and Todoist are projected
engagement surfaces. Full reference: docs/ARCHITECTURE.md.
Implementation Status
v0.1.0 shipped (2026-06-24) and is live on the box (Iris — persona, owned-core
brain/memory/planner, calendar + task read/write, confirmation + verify + quarantine + egress gates,
web research). v0.2.0 shipped (2026-07-04) — reliability + persona (D1–D5: live-state grounding,
memory re-arch + gated sensitive store, reconciliation-in-code, personality dials, cost slimming).
v0.2.1 (2026-07-18) is the cost-and-polish patch line (default Haiku 4.5 + owner /model switch,
PDF ingestion, time-grounding + task-hygiene fixes, extended-thinking re-enabled, model-eval harness).
The cost-reduction initiative continues on the pr-030 branch (bound session history, live on box,
close-out pending). See docs/0.2/ROADMAP.md for the ordered plan; full PR descriptions in prs/.
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 · 86 lines · 1,373 tokens per session scan A 9fa4fe255153
iris CLAUDE.md is an instructions file published in the GitHub repository rux-eth/iris (0 stars, last pushed 26d ago), licensed MIT. It adds 1,373 tokens to every session, about $0.0069 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
minimalist copilot-instructions.md
Instructions for DivyeshJayswal/minimalist, covering minimalist (coding tasks), persistence, fast path, bias and output.
omnifocus-claude-extension CLAUDE.md
Instructions for geoffdavis/omnifocus-claude-extension, covering claude.md, project overview, essential commands, build & development and install dependencies.
navi CLAUDE.md
Instructions for Affirm/navi, covering navi - claude code session monitor, setup instructions and removal instructions.
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
spec-kit AGENTS.md
AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.