decision-models
Skill1 starsChanged 43 days agoMITView on GitHub
Design, evaluate and wire JEV-style decision models (typed AI decisions with probabilities — TypeSafe Jev, Ollama's /v1/systemone, models like tev1 and jev-latest) into agent harnesses, skills and apps. Covers what they are, which decisions fit (routing, guards, injection flags, triage, verification) and which don't, a tested never-fail Python client with thresholds, an eval runner on labeled cases, and the optional-adapter pattern for skills and hooks. Use this whenever someone wants to add a classifier, gate, guardrail, router, prompt-injection check, yes/no or multiple-choice judgment to an agent, skill or workflow, asks "could a small/local model decide this instead of the LLM", mentions Jev, System One, TypeSafe, decision models, typed decisions or Ollama decision models — even if they don't name decision models explicitly.
- Reads credentials
2 other projects use a version of this.
Adopted by
Most recently taken up by
- timmo001/skillstheir own version, first seen Nov 7, 2026
- Haple/system-onetheir own version, first seen Oct 20, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- timmo001/skills★ 2same name
- Haple/system-one★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository3 other files it gives its agents.
- dev-workflowSkill
- feed-second-brainSkill
- okfSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 45
Before recommending anything, ask yourself honestly whether the decision model adds value over the LLM that is already running: calibrated …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Decision models
A decision model answers closed, typed questions about some text — yes/no, one of a list, a level on a rubric — and returns probabilities instead of prose. Your code applies thresholds and decides what to do. They are a small, fast, cheap complement to an LLM: the LLM writes, the decision model checks, sorts and warns.
This skill packages what was learned building real guards into an agent skill: what works, what failed, the safety rules, and code you can reuse.
The five rules
- A signal, never a verdict. The model sorts, warns, hints or asks for caution. The agent, the user or deterministic code makes the call.
- Only in the safe direction. Act on an answer only when it makes the system more careful (warn, ask, keep open, escalate). Never let it remove a check.
- Thresholds live in code. Below "sure" (default 0.85 / 0.15) the answer is
undecidedand the caller judges as if there were no model. - Never fail the caller. Server down, model missing, timeout, bad reply → "no signal", exit 0, work goes on. Check health once per run; stop calling after the first failure.
- Measure before shipping. Label 20–50 real cases per question and run the eval. Ship only questions with no sure-but-wrong answers (or an acceptable cost).
Why rule 2 matters so much: the text being judged is often written by outsiders (PR comments, tickets, web pages), and it can try to steer the judge too. A decision model can raise an alarm; it cannot clear text as safe.
How to work
Pick the path that matches the request:
| The user wants to… | Do this |
| understand decision models, compare to LLMs, pick a model | read references/what-are-decision-models.md and answer |
| know where a decision model would help in their skill, harness or app | read references/use-cases.md; read their code/steps; run each candidate through the fit test and the direction rule; recommend 2–3, say what you'd skip and why |
| call a decision model from code | read references/implementing.md; prefer scripts/jev.py with a spec file; write tests with a mock transport |
| add decision signals to a skill or hooks | read references/in-skills.md and follow its step-by-step and checklist |
| check whether a model is good enough | write a spec + labeled cases, run scripts/eval.py, report right / WRONG / undecided per question |
Before recommending anything, ask yourself honestly whether the decision model adds value over the LLM that is already running: calibrated numbers, the same answer every run, a log, privacy, or isolation from untrusted text. If none apply, say so.
Bundled files
| File | What it is |
scripts/jev.py | Never-fail CLI and library: health and ask --spec … --state-file/--state-json …; providers ollama (default, local) and typesafe (cloud, TYPESAFE_API_KEY). Prints a STATUS: block, always exits 0. |
scripts/eval.py | Runs labeled cases through the same thresholds; prints right / WRONG / undecided and MISSED flags. |
scripts/test_jev.py | Unit tests with a mock transport (no model needed): uv run scripts/test_jev.py. |
examples/pr-comment.spec.json | A worked spec: injection flag, general-rule hint, comment kind. |
examples/pr-comment.cases.json | 30 labeled PR comments for it (5 injections). |
Quick start (local):
ollama pull tev1
uv run "<skill_dir>/scripts/jev.py" health --model tev1
uv run "<skill_dir>/scripts/eval.py" --spec "<skill_dir>/examples/pr-comment.spec.json" --cases "<skill_dir>/examples/pr-comment.cases.json" --model tev1uv run installs the typesafe-sdk dependency declared inside each script. Suggest setup
commands to the user; don't install software, pull models or create keys yourself — they cost
disk, memory or money. Never put an API key in a command line, a file in the repo, or the chat.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.