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/alexshchuka/mirabilis/agents-mdgit clone --depth 1 https://github.com/AlexShchuka/mirabilisWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/alexshchuka/mirabilis/agents-md)<a href="https://agentmods.dev/instructions/alexshchuka/mirabilis/agents-md"><img src="https://agentmods.dev/badge/instructions/alexshchuka/mirabilis/agents-md.svg" alt="Measured on agentmods" height="20"></a>What 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 5.1 | $0.03582 | $0.03582 |
| Opus 5 | $0.01791 | $0.01791 |
| Sonnet 5 | $0.00716 | $0.00716 |
| Haiku 4.5 | $0.00358 | $0.00358 |
Grade A, and why
mirabilis 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 5d 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 — 174 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
mirabilis is a personal, open-source cross-platform dev container that runs Claude Code with
full autonomy, the neuro-matrix harness
preinstalled, persistent memory, and open egress. CLAUDE.md is a
symlink to this file. Setup: README.md. Threat model: SECURITY.md.
Boundaries
- The container is the security boundary. Inside it the agent has full freedom —
root,
sudo, any file. Behavioural limits (no push/force tomain, no credential exfiltration) are the harness's job, not the sandbox's. Sandbox provisions; harness behaves (G2). - Work lives in
/workspace, a named volume the sandbox owns.~/.claudeand~/.config/ghare persistent volumes (memory, auth)./tmpand everything else is ephemeral. - Egress is open: the container reaches the network directly — no in-container allowlist.
api.anthropic.comis reached via the chain: claude → headroom (:8787, observability + MCP) → host auth proxy (token injection) → Anthropic. The real Claude token never enters the sandbox — only a per-session key does. Stopping credential exfiltration is the harness's behavioural job.WebFetch/WebSearchgo via the Anthropic API and always work. Deliberate additional egress edge:internal/engine/localllmPOSTs tohost.docker.internal:1234(LM Studio default port) to offload prompts to a host-local model. This edge is intentional and hardcoded tohttp://host.docker.internal:1234/v1; it carries only prompt text, never the Claude token. If an egress allowlist is added in future, it must includehost.docker.internal:1234. - Secrets: single source of truth per platform — keychain (macOS) / file
0600(Linux/WSL). Entry namemirabilis-<key>(no doubled suffix). The Claude OAuth token stays on the host; only the per-session key reaches the container. The Telegram token is host-side; Telegram chat-id is written to.mirabilis/chat-id(not an env var).
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.
- 5d ago First seen · 174 lines · 3,582 tokens per session scan A b81156c3219e
mirabilis AGENTS.md is an instructions file published in the GitHub repository AlexShchuka/mirabilis (5 stars, last pushed 3d ago), licensed MIT. It adds 3,582 tokens to every session, about $0.0179 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
aicontainer AGENTS.md
AGENTS.md instructions for stefanoginella/aicontainer, covering agents.md, repository layout, image tag pinning (read before editing the cli), version drift / update notifications and project identity and host-side validation.
code-box CLAUDE.md
Claude Code instructions for nezhazheng/code-box, covering claude.md, project overview, auto-update architecture, key files for auto-update system and installation and usage.
aicontainer CLAUDE.md
Claude Code instructions for stefanoginella/aicontainer, a project described as: Sandboxed devcontainer for running Claude Code, Codex, and OpenCode in bypass / auto-approve mode.
ironcurtain CLAUDE.md
Claude Code instructions for provos/ironcurtain, covering claude.md, general workflow, git & worktrees, git workflow and platform considerations.
claude-code-settings copilot-instructions.md
Instructions for feiskyer/claude-code-settings, covering claude.md, environment setup, required dependencies, configuration and skills.
vibe-coding-prompt-template backend.instructions.md
Instructions for KhazP/vibe-coding-prompt-template: Read AGENTS.md, agentdocs/techstack.md, and agentdocs/codepatterns.md.