mimo-code

A workflow for delegating coding tasks to mimo, a command-line coding assistant, while another agent coordinates and reviews the work.

In plain words
What is it for?
Use it to start or resume mimo coding sessions, choose a provider and model when needed, and review the resulting code changes.
Why use it?
It moves long implementation sessions and their output away from the main conversation, making delegation and review easier to manage.

Skill for Claude CodeCodex

Install

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.

agentmods
npx agentmods add skills/onsails/skills/mimo-code
Any agent
npx skills add onsails/skills --skill mimo-code
Clone the repo
git clone --depth 1 https://github.com/onsails/skills

Made for: Claude Code, Codex.

Per session 75 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,338 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00075 $0.01338
Opus 5 $0.00037 $0.00669
Sonnet 5 $0.00015 $0.00268
Haiku 4.5 $0.00007 $0.00134

Measured 2d ago against content hash 50d30f45a5e4, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

mimo-code 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.

mimo-code/skills/mimo-code/SKILL.md · 86 lines

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.

mimo-code

Overview

Delegate a write-capable coding session to mimo, staying the conductor. Heavy work and tokens move to mimo; you orchestrate and review the diff.

When to use

  • The user invokes /mimo-code [provider/model] [variant] <task> or asks to run/resume mimo.
  • A prior mimo session needs continuation (it asked something or was cut off).

The rule that shapes everything

Never run mimo in the main thread. ALWAYS dispatch the mimo-delegate subagent (model sonnet) so the live NDJSON stream and any blocking stay out of the main context. You only ever receive its distilled summary.

Parse the invocation

Raw slash-command arguments: $ARGUMENTS

Interpret as [provider/model] [variant] <task> (empty when triggered by description match rather than /mimo-code — then read intent from the user's message):

  • A leading xxx/yyy token → model. A following minimal|low|medium|high|max → variant.
  • The rest is the task; resume/continue → resume the recorded session (see Resume). Natural-language requests map the same way.

Resolve the model — ASK, do not default (only when no model was given, fresh runs only)

  1. Gather options via the mimo-resolve subagent (model sonnet). It runs the resolve-models launcher and returns the authenticated provider∩model options and variants — keeping the catalogue output out of the conductor's context. Empty options (no authenticated provider) → STOP and tell the user to run mimo providers login.
  2. Using the returned options (the authenticated intersection — count models, not providers): Exactly one usable model → auto-pick (asking is pointless). ≤4 → one AskUserQuestion listing them. >4 (the common real case — a single provider can expose dozens) → narrow, never dump: AskUserQuestion the provider first (the authenticated set is small), then AskUserQuestion ≤4 models for that provider — lead with mimo-resolve's recommended id, and rely on the auto-added Other free-text for the long tail. One authenticated provider offering several models is still "more than one" → ASK; "low-risk"/"mechanical"/cost or proactivity never justify a silent pick, and printing the whole catalogue for the user to retype an id verbatim is the anti-pattern to avoid.
  3. Ask effort/variant with a ≤4 menu — AskUserQuestion caps at 4 options, so you cannot list all five variants + a default. Offer default (omit --variant) plus a few variants, each with a one-line description (default is the safe recommendation for a one-shot delegation); minimal/medium reach via Other.

Read the full file on GitHub · 86 lines

Changes

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.

  1. 2d ago First seen · 86 lines · 75 tokens per session scan A 50d30f45a5e4

Subscribe to this mod's changes

mimo-code is a skill published in the GitHub repository onsails/skills (43 stars, last pushed 7d ago), licensed MIT. It adds 75 tokens to every session and 1,338 once invoked, about $0.0004 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.