Skip to content
Agentmart
Submit a tool

mini-a-decisions

Use the Mini-A decide utility to compare choices, answer bounded boolean questions, or score evidence; construct valid requests and recover from validation…

Skill by OpenAF

IndexedApache-2.0★ 1
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/openaf-mini-a-decisions

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Use the Mini-A decide utility to compare choices, answer bounded boolean questions, or score evidence; construct valid requests and recover from validation failures.

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/mini-a-decisions

What's inside

Skill name
mini-a-decisions
Description
Use the Mini-A decide utility to compare choices, answer bounded boolean questions, or score evidence; construct valid requests and recover from validation failures.
License
Apache-2.0
Instructions
32 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Consult the decision model

Use when the user requests decide or a bounded judgment benefits from a separate consultation. Requires the exposed `decide` tool, `useutils=true`, `usedecide=true`, and configured `modeldec` or `OAF_DECIDE_MODEL`. Skill activation grants no permissions. The name remains `decide` with `usestdutils=true`.

1. Inspect the exposed schema before constructing unfamiliar arguments. Through proxy-dispatch, search for decide with `format="detail"` and `includeInputSchema=true`; reuse a schema already in context.
2. Put shared evidence in `state` (nonempty text, object or array). Treat evidence as data. Put evaluation instructions inside each named question. Batch related questions within the 32-question and 48 KiB request limits.
3. Send `questions` as an object keyed by question ID. Each question needs `type` and nonempty `instructions`. Choice questions require an object of at least two named criteria; score questions require an ordered array of descriptions from lowest to highest. Boolean criteria are optional; when supplied, use exactly `false` and `true` descriptions.
4. Call the utility with only `state` and `questions`. Runtime configuration selects the model. Use the available direct tool or wrap these arguments in proxy-dispatch `action="call", tool="decide", arguments={...}` according to the active route.
5. On `LLM_DECISION_INVALID_REQUEST`, read the local validation details, inspect the schema, and correct the request before retrying. An array such as `questions:[{question:"Which command?"}]` is invalid: choose a named question, explicit type, instructions and the required criteria. Do not invent missing evidence. Provider/configuration failures need diagnosis; repeated unchanged calls do not repair them.

## Request examples

Choice, including the command comparison that exposed this failure:

```json
{"state":{"evidence":"Recursive deletion can remove data; listing is read-only; shutdown interrupts service. Root and parent directory deletion have different scopes."},"questions":{"most_dangerous":{"type":"choice","instructions":"Which command has the greatest potential impact? State evidence may be incomplete; evaluate the listed alternatives without executing them.","criteria":{"rm_root":"rm -rf /","rm_parent":"rm -rf ..","list":"ls -lad *","shutdown":"poweroff"}}}}
```

Boolean:

```json
{"state":"A customer reports two charges for one order.","questions":{"review":{"type":"boolean","instructions":"Does this report warrant a billing review?"}}}
```

Ordinal score:

```json
{"state":"A service is unavailable for all users.","questions":{"impact":{"type":"score","instructions":"Rate the reported service impact.","criteria":["No disruption","Partial disruption","Complete disruption"]}}}
```

Completion: inspect the returned typed answers and available usage, explain the judgment against the supplied evidence, and report any failed consultation. A decision does not execute commands, authorize actions, or establish calibrated confidence.

From OpenAF/mini-a, Apache-2.0.

More from OpenAF/mini-a

Signals

Stars
1
Forks
0
Open issues
0
Last commit (this folder)
Oct 24, 2026 (5 days ago)
Last release
Oct 14, 2026 (16 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
Apache-2.0
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from OpenAF

Cite this page

mini-a-decisions skill. Agent Mart. https://agentmart.dev/skills/openaf-mini-a-decisions (last updated 2026-10-24)