jev-agent-skill-router
Confidence-aware skill router that uses Jev to pick agent skills from large catalogues, declining or asking for review on weak matches.
Overview
Jev Agent Skill Router is a Python CLI and library that routes a request to the agent skill it needs, motivated by a Hermes Agent profile exposing 443 skills. Given a request and a catalogue of skill names with public descriptions, it returns route, no_skill or review, a three-valued judgement of whether a specialist is needed with its Noul probability, the final Choice confidence and winner probability, and full evidence for every call including candidates, distributions, latency and token usage. Catalogues are sorted, split into deterministic batches and evaluated in parallel, with survivors compared in later rounds.
On a public benchmark of 24 synthetic skills and 72 synthetic requests, the README reports 68 of 72 routed correctly (94.4%) versus 51 of 72 for a lexical baseline, with no wrong skills or needless loads but three unnecessary reviews and one validation failure; median latency was 1,287 ms. It is candid that these are exploratory results on reused data after revising questions, and that typed output guarantees shape, not correctness. It is an optional layer, not a Hermes replacement.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Does this request need a specialist skill? | yes · no · uncertain |
| Choice | Which skill in this batch best fits the request? | — |
How it works
- Load the skill catalogue and split into deterministic batches
- Ask Jev whether a specialist is needed
- Run parallel Choice rounds over batches; keep top candidates
- Return route, no_skill or review with evidence
Strengths
- Declines or requests review instead of guessing
- Full per-call evidence and latency reported
- Outperformed a lexical baseline on its synthetic benchmark
Caveats
- Exploratory results on reused, synthetic data
- Median latency over a second due to multiple rounds
At a glance
- Type
- Agent skill
- Status
- research
- Language
- Python
- Runtime
- CLI, library
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
Agent infrastructure: routes agent skill selection through typed, confidence-aware Jev decisions so weak matches are declined instead of guessed.
yibie/awesome-jevTyped, confidence-aware agent skill routing with TypeSafe Jev.
hellogumbo/awesome-jevRoutes a request to the agent skills it needs, with a confidence floor below which it loads nothing.
MrJev/awesome-jev
Frequently asked questions
What does the router decide?
Whether a request needs a skill and, if so, which one, or whether to send it for review.
How accurate is it?
94.4% on 72 synthetic requests in the README's exploratory run.
Does it replace Hermes skill loading?
No, it is an optional routing layer.
Related projects
Opt-in Hermes Agent plugin that asks Jev to name the one installed skill worth loading before the model call.
Agent skillbetaPythonMIT plugin and local MCP server that hands an agent's closed questions to Jev or local Laya, with policy gates and receipts.
Agent skillproductionPythonClaude Code and Codex skill that routes a task to one of 27 specialist roles, with an optional Jev engine for catalog selection.
ProjectbetaPythonClaude Code hook that decides which agent or skill should handle a request using YAML rulebooks plus one Jev Choice with confidence bands.
HookbetaPythonOrchestrator that probes which coding agents are installed and logged in, then has Jev pick the best one for a task.
Agent skillbetaPythonHooks for Claude Code, Codex and pi that use Jev to suggest a skill each turn, plus a rule gate for subagent model routing.
HookbetaJavaScript