axis-method
Public home of AXIS-26 v1.0 — the Agentic eXecutable Intent Specification — and the reference plugins / adapters that make it work in every major AI coding tool.
Works with every AI coding tool
AXIS-26 principle 5 is tool-portable, vendor-replaceable, model-fungible. The spec artifacts (Constitution, AGENTS.md, routing.yaml, EARS specs, SKILL.md) are plain Markdown / YAML and load identically in any agent.
| Tool | Install | Coverage |
|---|
| Claude Code | /plugin marketplace add thaithienvanid/axis-method then /plugin install axis-core@axis-method | Full plugin: 7 slash commands, 6 skills, 3 subagents, 4 lifecycle hooks |
| Cursor | ./scripts/init.sh <repo> --cursor | Project rules + slash commands + git pre-commit for G2 |
| Codex CLI | ./scripts/init.sh <repo> (AGENTS.md auto-loaded) | AGENTS.md context + git pre-commit |
| Gemini CLI | ./scripts/init.sh <repo> --gemini | AGENTS.md + 7 TOML slash commands in ~/.gemini/commands/ |
| Aider | ./scripts/init.sh <repo> --aider | .aider.conf.yml + auto-lint wired to CWE scanner |
| Anything reading AGENTS.md | ./scripts/init.sh <repo> | Constitution + AGENTS.md + routing — full conformance is achievable |
The Markdown control plane under .axis/ is fully agent-agnostic. The conformance test (§3.4) checks repository state, not which agent edited the files.
See docs/PORTABILITY.md for the full matrix and per-tool notes.
Quick start (any tool, ~2 hours per AXIS-26 §1.2)
git clone https://github.com/thaithienvanid/axis-method.git
cd <your-project>
/path/to/axis-method/scripts/init.sh . # tool-agnostic core
# Or with adapter flags:
/path/to/axis-method/scripts/init.sh . --cursor # Cursor adapter
/path/to/axis-method/scripts/init.sh . --gemini # Gemini CLI adapter
/path/to/axis-method/scripts/init.sh . --aider # Aider adapter
/path/to/axis-method/scripts/init.sh . --git-hooks # G2 enforcement via Git pre-commit
After init, edit constitution.md and .axis/routing.yaml for your project, then run a first Change through Specify → Build → Verify → Deploy → Observe.
Claude Code one-liner
/plugin marketplace add thaithienvanid/axis-method
/plugin install axis-core@axis-method
What's in this repo
axis-method/
├── AXIS-26.md # the specification (CC BY 4.0)
├── axis-core/ # Claude Code reference plugin (Apache 2.0)
│ ├── .claude-plugin/plugin.json
│ ├── commands/ # 7 slash commands
│ ├── skills/ # 6 skills
│ ├── agents/ # 3 subagents
│ ├── hooks/ # 4 lifecycle hooks (POSIX bash, fail-soft)
│ └── templates/ # tool-agnostic templates
├── adapters/
│ ├── cursor/ # .cursor/rules/axis-26.mdc + slash commands
│ ├── codex/ # AGENTS.md auto-load + prompt one-shots
│ ├── gemini/ # ~/.gemini/commands/ TOMLs
│ └── aider/ # .aider.conf.yml + CONVENTIONS.md mapping
├── scripts/
│ └── init.sh # bootstrap any repo, any tool
├── docs/
│ └── PORTABILITY.md # multi-tool setup matrix
├── ap/ # AXIS Proposals (substantive spec changes)
├── .github/ISSUE_TEMPLATE/ # Bug, Question, Proposal seed
├── .claude-plugin/marketplace.json
├── README.md (this file)
├── CONTRIBUTING.md CODE_OF_CONDUCT.md CHANGELOG.md SECURITY.md
├── DISCLAIMER.md LICENSE
Free by default
The plugin and adapters require no paid third-party service. Optional vendor integrations (Linear, Braintrust, Snyk) are documented in axis-core/templates/mcp.example.json and are strictly opt-in. Free substitutes (Promptfoo, Semgrep, GitHub Projects) are listed alongside each.
Conformance tiers
Per AXIS-26 §3:
- Minimal (§3.1, M1–M5) — Constitution + AGENTS.md + routing.yaml + ≥1 skill + 1 Change end-to-end.
- Standard (§3.2, S1–S5) — Adds EARS form, brownfield delta protocol, eval thresholds enforced, DX Core 4 + MTTV tracking.
- Full (§3.3, F1–F5) — Adds cross-cutting context files, drift detection, AI-CFR tracking, quarterly Constitution Review, shared skills library.
upstash
wshobson
nyldn
greyhaven-ai
wshaddix
Jeffallan