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 skills/thgmatajs/squirrel-mode/initnpx skills add thgMatajs/squirrel-mode --skill initgit clone --depth 1 https://github.com/thgMatajs/squirrel-modeWhat 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 | $0.00035 | $0.02108 |
| Opus 5 | $0.00017 | $0.01054 |
| Sonnet 5 | $0.00007 | $0.00422 |
| Haiku 4.5 | $0.00003 | $0.00211 |
Grade A, and why
init 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 — 128 lines — stays where its author put it; the contents beside it link to each section on GitHub.
squirrel-mode calibration
/squirrel:init builds the user's personal squirrel-mode profile through a seven-question interview. Follow this procedure exactly, in every session it runs in.
Rules for the whole interview
- Ask exactly one question per message. Never combine two questions in the same message, and never ask a follow-up question in the same message as the question itself.
- Every question is multiple-choice, with two to four options labeled A, B, C, D (only as many as the question needs, never more than four), plus a final line offering "or type your own answer". When the reply is a free-form answer rather than a lettered option, map it to the closest listed option by meaning, and state which one you mapped it to in one line, before continuing. This applies to every question in this interview, not only question 2 - question 2's own mapping table below is the one case detailed enough to need its own worked example, not a special case.
- Show progress on every question in the exact form
Question N of 7(for exampleQuestion 3 of 7), where N is the number of the current question. - Wait for the user's reply before asking the next question.
- If a reply matches neither a lettered option nor a plausible free-form answer, ask the same question again once, still showing the same
Question N of 7line. - Before the profile exists, mirror the language the user is currently writing in for the questions and the confirmation. This interview is what creates
language; it cannot yet read a value that does not exist. - If the user goes off-script mid-interview - asks something unrelated, reports a problem, or otherwise does not answer the current question - address what they raised first, then ask whether to resume from the same question you were on before that happened. Do not silently drop the interview, and do not resume without asking.
The seven questions, in order
Question 1 of 7: Language
Ask which language responses should use.
A. Portuguese (pt-BR) B. English (en) C. Spanish (es) D. Match whatever language I write in (auto)
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 · 128 lines · 35 tokens per session scan A 4755e649b268
init is a skill published in the GitHub repository thgMatajs/squirrel-mode (2 stars, last pushed 13d ago), licensed MIT. It adds 35 tokens to every session and 2,108 once invoked, about $0.0002 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 skills, from other repositories
adhd-and-47-tabs
Use when a user needs a direct, low-friction response for research, studying, writing, planning, decisions, administrative work, troubleshooting, or a multi-turn project, especially when they seem overwhelmed, distracted, stuck starting, interrupted, burdened by too many options, or likely to lose the active thread.
hyperfocus
ADHD-friendly output formatting. Restructures responses with chunking, visual hierarchy, and front-loaded key points. Modes: clean, flow (default), zen. Use when user says "hyperfocus", "focus mode", "adhd mode", "adhd friendly", or invokes /hyperfocus.
task-anchor
Version: 1.0 Activation: MANDATORY on every session start Persona: Executive Function Proxy (firm, non-judgmental, binary choices only).
workflow
Professional AI programming assistant with structured workflow (Research -> Ideate -> Plan -> Execute -> Optimize -> Review) for developers.
r3f-animation
React Three Fiber animation - useFrame, useAnimations, spring physics, keyframes. Use when animating objects, playing GLTF animations, creating procedural motion, or implementing physics-based movement.
r3f-best-practices
React Three Fiber (R3F) and Poimandres ecosystem best practices. Use when writing, reviewing, or optimizing R3F code. Triggers on tasks involving @react-three/fiber, @react-three/drei, zustand, @react-three/postprocessing, @react-three/rapier, or leva.