hermes-jev-north-star
Hermes skill that turns an intention into a judgeable finish line and a run prompt, with Jev deciding fuzzy requirements at the gate.
Overview
north-star is a Hermes Agent skill that turns an intention into a finish line an agent can be held to. Long autonomous runs fail by stopping early and calling it done or by never stopping; both stem from lacking a written finish line with checkable states. The skill interviews the owner until one exists, refuses if the result cannot be judged, and produces three artifacts: the star, the generated prompt that starts the run, and the gate that decides whether the work is finished.
The gate has two layers: machine-checkable parts run as plain shell commands, and the parts no script can decide, such as whether a stranger completed checkout without contacting support, go to Jev, which answers with a probability and confidence so the gate stays code and never reads a model's opinion of its own progress. Everything else runs offline in dependency-free Python; Jev is the default judge when a key is present, and the gate says plainly when it is not. It is unofficial.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is this fuzzy requirement of the finish line actually met? | yes · no |
How it works
- Interview the owner to define the finish line
- Generate the run prompt
- Run shell checks for machine-checkable parts
- Ask Jev about fuzzy requirements; gate passes only if all are met
Strengths
- Prevents premature or endless runs
- Gate stays deterministic code
- Works offline without a key for the deterministic half
Caveats
- Requires cloning for scripts in addition to skill install
- Fuzzy checks need a TypeSafe key
Getting started
hermes skills install https://raw.githubusercontent.com/poponline63/north-star/main/SKILL.md, then clone the repo for scripts and run /north-star <goal>.At a glance
- Type
- Agent skill
- Status
- beta
- Language
- Python
- Runtime
- Hermes Agent skill + scripts
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
A Hermes goal-checking skill that saves requirements, creates a run prompt and checks completion evidence.
logicrw/awesome-jev-projects
Frequently asked questions
What does Jev decide?
Whether fuzzy requirements of the finish line are met.
Does the agent judge its own progress?
No, the gate is code using shell checks and Jev probabilities.
Do I need a key?
Only for the fuzzy checks; the deterministic half runs offline.
Related projects
Typed judgment layer for Claude Code, Codex and Hermes Agent: deterministic facts in code, language judgments in Jev, with receipts.
ProjectbetaPythonHermes Agent skill that turns an intention into a checkable finish line, with shell checks and Jev judging the fuzzy requirements.
Agent skillprototypePythonMCP server giving coding agents fast typed decisions (ship gates, risk checks, file triage, page checks) as accept, review or escalate.
MCP serverbetaTypeScriptIntent compiler skill that turns vague coding requests into a typed IntentSpec, with Jev or LAYA as optional decision backends.
ProjectbetaPortable skill pack giving agents a conservative Jev second opinion on intent, completion, routing and bounded browser navigation.
ProjectbetaPythonCodex plugin that consults Jev to choose tools, models and skills, diagnose failed commands and check completion claims.
PluginbetaTypeScript