Epistemic Protocols plugin
Plugins for Claude Code and Codex bring key decisions to the human before coding: /inquire, /elicit, /bound and ten more.
- 176 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The installer is downloaded and run from the network and the plugins add commands and hooks to the session.
Install
- Works in
Manual install
curl -fsSL https://raw.githubusercontent.com/jongwony/epistemic-protocols/main/scripts/install.sh | bashInstalls all protocols together with the `route` and `epistemic-cooperative` plugins. The script is downloaded and run from the network.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The installer is downloaded and run from the network and the plugins add commands and hooks to the session.
This is third-party code. Review the repository files before installing.
What it does
Epistemic Protocols is a set of 13 protocols for human and AI collaboration. Each runs through its own slash command at a point where work would otherwise be redone: `/inquire` finds what is unknown and what is confirmed, `/elicit` helps state intent, `/bound` shows which decisions to make yourself, `/preview` and `/sketch` let you see options, `/ground` checks a carried-over analogy, `/apportion` cuts a goal into units before an autonomous run, `/recollect` finds a past discussion and `/grasp` helps understand code. A separate `epistemic-cooperative` plugin adds `/onboard` to pick a protocol from recent sessions, `/probe`, `/gate-check`, `/white-bear` and `/zero-shot`. An experimental `route` plugin picks a suitable protocol from context. A protocol contract is also written as a Lean model, and each protocol is its own plugin.
Who it is for. People who hand large tasks to agents and tire of redoing results because a wrong direction went unnoticed at the planning stage.
Good fit when
- Before handing work to an agent you must see what it does not know and which decisions stay with you.
- The intent is vague and key decisions surface only through clarifying questions.
- One goal is cut into units for an autonomous run with completion conditions for each.
- A discussion from a month ago is vaguely remembered and has to be found across sessions.
Not a fit when
- The task is small and fits one precise request: the procedure adds extra questions.
- The agent runs unattended, for example through `codex exec`: decisions arrive after the work has started.
- You need stability on another agent: behavior outside Claude Code is called provisional.
Example request
Run `/inquire` before I hand you preparation for the churn meeting and return what is confirmed, what conflicts and what only I decide.Limitations
Protocols work through checkpoints that present decisions and their consequences but do not by themselves suspend execution: on a non-interactive run such as `codex exec` decisions may arrive after work has started. Behavior in agents other than Claude Code is called provisional pending accumulated use. The `route` plugin is experimental and its wording and hook set change between releases. The installer is downloaded and run from the network, so read the script before running.
How to disable. Disable plugins with `claude plugin disable <name>@epistemic-protocols` or remove them through `/plugin`. For Codex remove the marketplace with `codex plugin` commands.
FAQ
How do I install Epistemic Protocols?
For Claude Code run `curl -fsSL https://raw.githubusercontent.com/jongwony/epistemic-protocols/main/scripts/install.sh | bash`: it installs all protocols together with the `route` and `epistemic-cooperative` plugins. Codex has `scripts/install-codex.sh`.
Which commands do the protocols have?
`/inquire`, `/elicit`, `/ideate`, `/preview`, `/sketch`, `/ground`, `/induce`, `/apportion`, `/contextualize`, `/bound`, `/recollect`, `/grasp` and `/conduct`. You pick the protocol by what is missing right now.
How do I turn off route or epistemic-cooperative?
After installing run `claude plugin disable route@epistemic-protocols` or `claude plugin disable epistemic-cooperative@epistemic-protocols`. The protocols stay callable by their own commands.
What does the /onboard command do?
It gives a quick recommendation from your recent sessions and can continue into guided learning with scenarios, trials and quizzes. It is part of the `epistemic-cooperative` plugin.
What do Lean and static checks cover?
Lean checks the stated properties of a protocol model under its assumptions. `/verify` checks file structure and references. They do not measure whether protocols reduce rework.
Discussion
No questions or reviews yet.
Related
PluginsA Claude Code plugin that hands coding tasks to Codex agents, monitors them and verifies the results independently
A Claude Code plugin: an orchestrator and 12 specialists run Prepare, Architect, Code, Test with memory and a session journal
Multi-agent orchestration for Claude Code: teams, roles, autopilot and task routing across skills and subagents
claude-mods adds 112 skills, 3 agents and 3 commands to Claude Code: stacks, agent operation, defense, session continuity.
AI Builder Club Skills
AI Builder Club — Skills
A Claude Code plugin marketplace to make a repo agent-ready and run compounding agent loops
Workflows for Claude Code: /do, /omo, /bmad-pilot and more commands, with Codex, Gemini or OpenCode writing the code through codeagent-wrapper
README in short
The repository serves as a marketplace for both Claude Code and Codex: each protocol is its own plugin and utilities live in `epistemic-cooperative`. Collaboration principles are described in the `premise/` directory.