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/aetherwing-io/mish/claude-mdgit clone --depth 1 https://github.com/aetherwing-io/mishWrote 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/aetherwing-io/mish/claude-md)<a href="https://agentmods.dev/instructions/aetherwing-io/mish/claude-md"><img src="https://agentmods.dev/badge/instructions/aetherwing-io/mish/claude-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 | $0.01382 | $0.01382 |
| Opus 5 | $0.00691 | $0.00691 |
| Sonnet 5 | $0.00276 | $0.00276 |
| Haiku 4.5 | $0.00138 | $0.00138 |
Grade A, and why
mish 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 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 — 109 lines — stays where its author put it; the contents beside it link to each section on GitHub.
mish
LLM-native shell. A Rust binary that is both a CLI command proxy and a full MCP server for process supervision. It wraps shell commands with structured, context-efficient responses, manages concurrent PTY sessions, and provides process lifecycle control — all through a single binary.
Status
Design/spec phase. No source code yet. All work is in docs/.
Naming
The project is mish. Never use "llmsh" — that was the old name.
Docs
Start here: docs/ARCHITECTURE.md is the unified execution model — it shows how both modes (CLI proxy and MCP server) share a four-layer architecture: Entry Points → Category Router → Mode-Aware Handlers → Shared Primitives.
| Doc | Covers |
|---|---|
ARCHITECTURE.md |
Unified execution model, module structure, PTY risk assessment |
SPEC.md |
Project overview and quick reference |
mish_spec.md |
MCP server mode definitive spec (process table, yield, handoff, policy) |
PROXY.md |
CLI proxy mode and the 6-category routing system |
GRAMMARS.md |
Grammar schema, tool grammar specs, dialect system |
CLASSIFIER.md |
Three-tier classification engine |
STREAMING.md |
Streaming buffer, PTY capture |
DEDUP.md |
Deduplication engine |
PREFLIGHT.md |
Argument injection (quiet/verbose flags) |
VERBOSITY.md |
Verbose flag injection, file stat enrichment |
ENRICH.md |
Error enrichment (failure diagnostics) |
Three Modes, One Binary
Daemon (mish daemon)
Shared process table over Unix socket. All mish serve instances auto-connect to the daemon if running, enabling cross-agent visibility and coordination.
# Start daemon (background)
mish daemon &
# All subsequent `mish serve` instances proxy to the daemon
# Agents can see each other's processes via sh_session(action="list")
- Socket:
~/.local/share/mish/mish.sock - Proc log:
~/.local/share/mish/procs.jsonl(crash recovery) - Orphan re-adoption: on startup, checks PIDs from log, re-adopts survivors
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 · 109 lines · 1,382 tokens per session scan A bdf459bab631
mish CLAUDE.md is an instructions file published in the GitHub repository aetherwing-io/mish (5 stars, last pushed 5mo ago), licensed MIT. It adds 1,382 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
avibe AGENTS.md
AGENTS.md instructions for avibe-bot/avibe, covering agent guidelines for avibe, 1. project overview, 2. design philosophy and architecture, product and technical decision-making and core rule: fix at the highest appropriate layer.
Vibelution AGENTS.md
Instructions for CCDawn/Vibelution, covering vibelution agent rules, 1. identity and priority, 2. global red lines, 3. start and routing and 3.0 每任务先 router.
mcp-deep-search CLAUDE.md
Instructions for just-every/mcp-deep-search, covering claude.md, project overview, core modules & files, commands and development.
graylog-mcp CLAUDE.md
Claude Code instructions for jperelli/graylog-mcp, covering claude.md, what this is, layout, run / debug and configuration model.
memoryClaw CLAUDE.md
Instructions for tostechbr/memoryClaw, covering memoryclaw, project status, ✅ resolved: multi-user isolation (sprint 0 — 2026-03-06), next steps (sprint 2 focus) and sprint 0: multi-user isolation ✅ completa (2026-03-06).
avibe CLAUDE.md
Claude Code instructions for avibe-bot/avibe, a project described as: The local-first Agent OS — your AI partner lives on your own machine. Drive the official Claude Code, Codex & OpenCode from your browser or any chat app.