omp-typesafe
Jev-powered adversarial or advisory reviewer for the omp coding agent, plus a typesafe_ask tool, plan/stop gates and context compaction.
Overview
omp-typesafe is a reviewer extension for the omp coding agent that follows omp's advisor pattern. In its default adversarial role it argues that the last action or claim is wrong, unverified or incomplete; in advisory mode it suggests worth-checking steps, simpler alternatives, missed updates and ambiguities. Review notes are advice and never block a tool call.
A few pieces can intervene: an ambiguity gate can refuse a plan submission in plan mode until the user has been asked, an optional stop gate can ask the agent to keep going, and pipeline features for omp-skills users can block specific tool calls or take a Jev second opinion where exact checks cannot decide. It also registers a typesafe_ask tool exposing noul, choice and score directly, and can cut stale tool output from large contexts using Jev scores. The README warns that transcript parts and diffs are sent to TypeSafe and documents what leaves the machine. Without a key it stays inactive.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is the last action or claim wrong, unverified or incomplete?(inferred) | yes · no |
| Yes / no | Is this plan ambiguous enough that the user should be asked first?(inferred) | yes · no |
| Yes / no | Should the agent keep going rather than stop?(inferred) | yes · no |
How it works
- Watch the session as it unfolds.
- Ask Jev a battery of review questions for the configured role.
- Deliver deduplicated notes within budgets; gates intervene only in specific cases.
- Optionally compact stale tool output.
Strengths
- Second opinion without blocking normal tool calls.
- Direct typesafe_ask tool for the agent.
- Clear documentation of data sent to TypeSafe.
Caveats
- Sends parts of transcripts and diffs to TypeSafe.
- Adversarial notes may add noise.
Getting started
omp plugin install github:siddicky/omp-typesafe and export TYPESAFE_API_KEY.At a glance
- Type
- Project
- Status
- beta
- Language
- TypeScript
- Runtime
- omp extension, Bun
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
TypeSafe AI (Jev) adversarial reviewer and typesafe_ask tool for the omp coding agent.
hellogumbo/awesome-jev
Frequently asked questions
Does the reviewer block tool calls?
No, review notes are advice; only the plan ambiguity gate, stop gate and specific pipeline checks can intervene.
What does typesafe_ask do?
It lets the agent ask Jev noul, choice and score questions directly.
What data leaves my machine?
Everything the reviewer judges, including transcript parts and diffs, goes to TypeSafe.
Related projects
DeepSeek Harness plugin where Jev prunes long tool output, screens fetched pages for injected instructions and gates completion claims.
Agent skillbetaTypeScriptHermes plugin that asks Jev to review risky plans, check claims against sources, and judge whether a task is really finished.
PluginbetaClaude Code plugin with nine Jev-judged gates for rules, scope, intent, completion and claims that escalate but never approve.
HookbetaTypeScriptCalibrated alignment verifier that checks LLM responses and agent plans for sycophancy, deception, overreach and dark patterns with Jev.
ProjectprototypeTypeScriptClaude Code plugin sending small judgments to Jev: rule checks that block edits, verbatim compaction, subagent sizing and routing hints.
PluginbetaPythonClaude Code Stop hook that checks an assistant's claims against what it actually read this session using Jev, blocking unsupported ones.
HookbetaPython