hermes-jev-approvals
Hermes Agent approvals provider where one Jev request answers six typed questions and code derives APPROVE, DENY or ESCALATE.
Overview
hermes-jev-approvals replaces only the reviewer behind Hermes Agent's smart command approvals. Hermes already detects commands that need review and asks an auxiliary model for APPROVE, DENY or ESCALATE; this plugin serves that task with Jev, returns typed probabilities instead of prose, refuses every other task and registers no hooks.
One request asks six questions in parallel: a verdict Choice, a Noul for whether the configured smart policy explicitly allows the command, a blast-radius Score, and Nouls for whether the command text argues for its own approval, reads secrets, or sends content outbound. Deterministic code then applies ordered rules, such as escalating self-advocating commands, denying secret reads combined with outbound transmission, and escalating secret reads, before falling back to the model verdict. Malformed answers make Hermes escalate. According to list summaries, on 153 real commands it was 8.7x faster with 4.4x fewer user prompts; it is described as a proof of concept.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | verdict | APPROVE · DENY · ESCALATE |
| Yes / no | self_advocating: does the command text argue for its own approval? | yes · no |
| Yes / no | reads_secrets / sends_outbound | yes · no |
| Score | blast_radius | 0 trivial · 1 annoying · 2 severe recovery |
How it works
- Hermes flags a command for review
- One Jev request answers six typed questions
- Ordered code rules derive the verdict
- Malformed answers escalate to the user
Strengths
- Reported faster decisions and fewer user prompts
- Defends against self-advocating command text
- No hooks; approvals only
Caveats
- Proof of concept
- Requires Hermes smart approvals mode
Getting started
hermes plugins install anpicasso/hermes-jev-approvals/plugin, hermes auth add typesafe-jev, then set approvals.mode smart and auxiliary.approval.provider typesafe-jev.At a glance
- Type
- Hook
- Status
- prototype
- Language
- Python
- Runtime
- Hermes Agent approvals provider
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Agent approvals: proof of concept that puts Jev in front of Hermes Agent's command approvals, reporting 8.7x faster decisions and 4.4x fewer prompts to the user.
yibie/awesome-jevAn experimental Hermes command-approval provider for the auxiliary.approval task only.
logicrw/awesome-jev-projectsPoC: TypeSafe Jev as the reviewer for Hermes Agent smart command approvals. 8.7x faster, 4.4x fewer prompts, measured on 153 real commands. Approvals only.
hellogumbo/awesome-jevApprovals provider for Hermes Agent: it judges shell commands and refuses every other task, registering no hooks.
MrJev/awesome-jev
Frequently asked questions
What does it replace?
Only the auxiliary model reviewing Hermes smart approvals.
Who makes the final verdict?
Deterministic code applying ordered rules to Jev's answers.
How much faster is it?
List summaries report 8.7x faster with 4.4x fewer prompts on 153 commands.
Related projects
Safety gate plugin for Claude Code, Codex CLI and Antigravity: local sensitive-file checks, then Jev judges blast radius.
PluginbetaTypeScriptClaude Code mod that answers auto mode's tool-call permission question with eight Jev yes/no checks behind a shell-aware blocklist.
ProjectbetaTypeScriptRing Zero: kernel-level runtime security for AI agents, with Jev as an optional check that can only raise severity, never permit.
ServicebetaClaude Code hook that blocks known key formats locally and sends unknown high-entropy strings to Jev only in masked form.
HookbetaPythonClaude Code hook where Jev reviews each tool call against risk, authorization and scope questions and a local policy returns allow or deny.
ProjectbetaGoOpenCode 2 plugin that asks Jev six questions about every shell command (local or remote via FarHand) and prompts unless clearly safe.
MCP serverbetaTypeScript