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 skills add ArcticFox2029/chamnan --skill decidegit clone --depth 1 https://github.com/ArcticFox2029/chamnanWrote 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/skills/arcticfox2029/chamnan/decide)<a href="https://agentmods.dev/skills/arcticfox2029/chamnan/decide"><img src="https://agentmods.dev/badge/skills/arcticfox2029/chamnan/decide/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/arcticfox2029/chamnan/decide"><img src="https://agentmods.dev/badge/skills/arcticfox2029/chamnan/decide.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00093 | $0.02172 |
| Opus 5.5 | $0.00037 | $0.00869 |
| Sonnet 5.5 | $0.00019 | $0.00434 |
| Haiku 4.5 | $0.00009 | $0.00217 |
Grade A, and why
decide 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 23d 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 — 172 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Decide from counted evidence, not from an impression
Code calculates. A model reasons. This decides. The pattern is borrowed openly from the decision models now being built for agent systems, and one thing is deliberately different: the confidence here does not come from a model. It comes from evidence you counted. Nothing in this procedure calls out, and nothing about it needs a network.
That difference is the whole reason it can be trusted offline — and the reason step 3 is the one that cannot be skipped.
1 · Name the decision, and what being wrong costs
Write both before looking at anything. The cost sets the threshold; there is no global one.
"Is this file safe to edit?" Being wrong means the edit is silently discarded. "Does this component already exist?" Being wrong means a wasted afternoon, recoverable. "Is this a credential?" Being wrong once is the whole product failing.
One threshold per action, scaled to that cost. A single number for the whole system is how a gate ends up too strict for cheap mistakes and too loose for expensive ones.
2 · Write the questions before you look
Each question gets a type, and the type is one of three:
| type | shape | example |
|---|---|---|
| choice | one value from a set you named | generated · hand-written · cannot tell |
| count | a number, with what it is out of | 26 of 28 planted credentials caught |
| yes/no | with the evidence that decides it | is a test command in this session's log? |
Never a free-prose answer. Prose cannot be thresholded, cannot be compared to last week's, and cannot be checked by anybody else. If a question can only be answered in prose, it is not a decision — it is a discussion, and it belongs with a person.
3 · Answer each from something you counted — and record the count
This is the step that makes the rest honest.
A number you did not count is an impression wearing a number's clothes.
Where the counts come from, offline:
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.
- 23d ago First seen · 172 lines · 93 tokens per session scan A 67338edfd3b1
decide is a skill published in the GitHub repository ArcticFox2029/chamnan (10 stars, last pushed 4d ago), licensed MIT. It adds 93 tokens to every session and 2,172 once invoked, about $0.0004 per session on Opus 5.5. 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 2026-09-23.
Other skills, from other repositories
cloudflare-workers-testing
Comprehensive testing guide for Cloudflare Workers using Vitest and @cloudflare/vitest-pool-workers. Use for test setup, binding mocks (D1/KV/R2/DO), integration tests, or encountering test failures, mock errors, coverage issues.
api-testing
HTTP API testing for TypeScript (Supertest) and Python (httpx, pytest). Test REST APIs, GraphQL, request/response validation, authentication, and error handling.
bun-jest-migration
Use when migrating from Jest to Bun's test runner, import compatibility, mocks, and config.
bun-test-basics
Use for bun:test syntax, assertions, describe/it, test.skip/only/each, and basic patterns.
bun-test-lifecycle
Use for test lifecycle hooks: beforeAll, afterAll, beforeEach, afterEach, fixtures, preload.
bun-test-mocking
Use for mock functions in Bun tests, spyOn, mock.module, implementations, and test doubles.