TypeSafe skill router
Opt-in Hermes Agent plugin that asks Jev to name the one installed skill worth loading before the model call.
Overview
TypeSafe skill router is a Hermes Agent plugin for setups with hundreds of installed skills, where the model tends to read past the one that would do the job. Before the model call, it sends the request to Jev; when a skill genuinely fits, it appends a single skill_relevance line naming it to the user message, with a note to ignore it if it does not fit. When nothing fits, it injects nothing.
It uses a two-stage gate: a first request makes three judgments about the request, and if their mean is below a gate (0.30 by default) no second request is spent; otherwise the winner's own "does this skill do the specific thing asked" judgment must pass a fits threshold (0.40) with a margin. It is off by default, standard-library Python only, and every failure path logs one line and returns nothing so routing can never break a turn. The lists cite about $0.001 per routed turn.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Does this request call for a specialized skill at all?(inferred) | yes · no |
| Choice | Which installed skill is most relevant to this request? | — |
| Yes / no | Does this skill do the specific thing asked for? | yes · no |
How it works
- A user message arrives in Hermes.
- First request: three judgments; below the gate, stop.
- Second request: pick a skill and check it fits above threshold and margin.
- Append one advisory line naming the skill, or nothing.
Strengths
- Opt-in and fails silently without breaking turns.
- Saves the second request when the gate is not cleared.
- No dependencies; about $0.001 per routed turn per the list.
Caveats
- Specific to Hermes Agent 0.21+.
- Requires a TypeSafe key.
Getting started
hermes plugins install typesafe-skill-router, add TYPESAFE_API_KEY to ~/.hermes/.env, then enable it.At a glance
- Type
- Agent skill
- Status
- beta
- Language
- Python
- Runtime
- Hermes Agent plugin
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
An opt-in Hermes Agent plugin that asks Jev to suggest one relevant skill before a model call.
logicrw/awesome-jev-projectsTypeSafe (Jev) skill routing for Hermes Agent: names the one skill worth loading, before the model call. Opt-in, stdlib only, ~$0.001 per routed turn.
hellogumbo/awesome-jevHermes Agent plugin that names the one skill worth loading before the model call. Off by default, injects nothing when nothing fits, and does not spend its second request when the first gate is not cleared.
MrJev/awesome-jev
Frequently asked questions
What does it add to the conversation?
At most one line naming a relevant skill; nothing if no skill fits.
Is it on by default?
No, enabled is false until you switch it on.
What happens on errors?
It logs one line and returns nothing, so the turn continues normally.
Related projects
Confidence-aware skill router that uses Jev to pick agent skills from large catalogues, declining or asking for review on weak matches.
Agent skillresearchPythonMIT 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