Axiom

Think before AI thinks.

Focused workflow skills for Codex and Claude Code that turn requests into clear, actionable work.

Install | Update | Use | Skills | Documentation | Changelog

If Axiom helps your agent work, consider starring the repository. Stars help other Codex and Claude Code users find it.

---

Skills

Describe the outcome you want, or name a skill. Axiom selects a bundled skill from your host's skill list by its description; a request that matches none continues through the host as usual. Skills use the host's existing tools and permissions, with no background service or automatic updater.

Skill	What it does
using-axiom	Route installed Axiom skills from a session bootstrap
clarify-intent	Resolve unclear requests with options or a custom answer
delegate-simple-task	Delegate simple work to your ordered candidate models
task-planning	Create and revise plans around current requirements
continuous-progress	Supervised progress on a native scheduler
local-web-search	Constrain local searches and stop at site challenges
agents-architect	Initialize, maintain, and validate scoped AGENTS.md systems
agent-plugin-architect	Design and audit packaged plugin architecture
optimize-agent-usage	Reduce agent context and workflow overhead
review-axiom-task	Audit Axiom decisions from observable evidence
verify-completion	Check completion claims against current evidence
confirm-external-action	Bind and verify consequential external actions
traceable-git-submit	Trace checkpoints and explicit Git submissions
reversible-system-change	Plan and verify reversible system changes

The Skill Reference gives each skill's full description, a copyable example, and rename notices.

Installation

Run these commands in a terminal for your host.

Codex:

codex plugin marketplace add wheakerd/axiom --sparse .agents/plugins
codex plugin add axiom@axiom

Claude Code:

claude plugin marketplace add wheakerd/axiom
claude plugin install axiom@axiom

Each marketplace selects an immutable published version. A new version reaches both marketplaces only after its release is published and verified; development commits on main do not advance that selection.

In Codex, start a new session and trust the Hook in /hooks: Codex skips a new or changed Hook until you do, and Axiom's routing begins after that step. Claude Code has no separate Hook trust step and runs the Hook at the next SessionStart after the plugin loads. On Windows, Claude Code needs Git Bash for the Hook, and neither host has a recorded Windows observation yet; see Compatibility. Verifying an installed Hook against its checked-in definition is an optional reviewer step described in the Trust Model. The Getting Started guide owns the full installation and first-use procedure.

Updating

Refresh the installed marketplace snapshot from a terminal.

Codex:

codex plugin marketplace upgrade axiom

Claude Code:

claude plugin marketplace update axiom
claude plugin update axiom@axiom

After updating, start a new session; Codex holds the changed Hook until you trust it again in /hooks. Automatic update behavior is controlled by each host and your settings. See Managing an Installation for in-app updates, disabling, removal, and troubleshooting. For a marketplace clone timeout in either host, follow the guide's recovery procedure.

Usage

To name a skill, include $skill-name in Codex or type /axiom:skill-name in Claude Code. Try a read-only request that matches a skill:

Use the agents-architect skill to audit this repository's AGENTS.md
instruction system. Report findings only; do not modify files.

Then try an ordinary request:

Summarize the purpose of this README. Do not modify files.

The first request should select agents-architect; the second should continue normally without another skill. Any explanation of the routing decision is optional and follows the invocation notes; a quiet response is valid. These are expected behaviors, not a claim that your installation has already reproduced them.

Important

A request that matches no Axiom skill still follows your host's permissions and repository instructions. No match is not a safety certification. Selecting a skill does not grant permission to edit, publish, deploy, or delete anything.

Examples shows more requests and their expected behavior.

Star History

The Record star history workflow re-renders the chart daily from stars.csv on the stats branch; see the stats branch policy.

Documentation and Support

• Documentation index: guides and references by task.
• Skill Reference: bundled skills and examples.
• Latest release and Changelog: what changed in each version and any required action.
• Compatibility: supported hosts and known limitations, backed by the release-status record.
• Trust Model: security boundaries. Axiom is a public beta; its instructions do not replace host permissions or provide a sandbox.
• GitHub Discussions: questions and early ideas.
• GitHub Issues: bug reports, feature requests, routing cases, and compatibility reports.
• SECURITY.md: report a vulnerability privately.
• Contributing and Code of Conduct: contribution guidance and community expectations.

License

MIT.