agentile
v0.16.1Agentile for Claude — the Claude Code implementation of the Agentile methodology for 1–5 person teams who direct AI agents: capture → shape → spec → plan → build → verify → ship → learn. Ships the fixed loop machinery (skills, agents, hooks); each project tailors its content via scaffolded .agentile/ config.
By Keith RowellLicense: MIT0 GitHub starsUpdated last week
Directory evidence
- Runtimes
- Claude Code
- Parsed components
- 20 skill or MCP entries
- Source updated
- Sep 17, 2026
- Manifest status
- Canonical path parsed
The directory validates manifest shape and source location. It does not execute the plugin or provide a security endorsement. Review the indexing methodology →
Install agentile for Claude Code
claude plugin marketplace add IchenDEV/agent-plugin-mkt
claude plugin marketplace update agent-plugin-marketplace
claude plugin install agentile@agent-plugin-marketplacePaste and run these commands in a terminal with Claude Code. They add and refresh the PluginsMP catalog, then install this plugin.
The installer fetches third-party code from the source repository shown on this page. This directory validates manifest structure and source location, but does not perform a security audit; review the manifest, components, and source before installing.
Get the source manually
git clone https://github.com/agenta-consulting/agentileClone the source repository, then follow its setup instructions to add the plugin to a compatible client. The repository root is the plugin root.
Plugin files
├── .claude-plugin/plugin.json├── skills/ag-abandon/SKILL.md├── skills/ag-build/SKILL.md├── skills/ag-capture/SKILL.md├── skills/ag-customise/SKILL.md├── skills/ag-decompose/SKILL.md├── skills/ag-deploy/SKILL.md├── skills/ag-inbox/SKILL.md├── skills/ag-init/SKILL.md├── skills/ag-loop/SKILL.md├── skills/ag-map/SKILL.md├── skills/ag-new-project/SKILL.md├── skills/ag-next/SKILL.md├── skills/ag-outcome/SKILL.md├── skills/ag-plan/SKILL.md├── skills/ag-prioritise/SKILL.md├── skills/ag-retro/SKILL.md├── skills/ag-shape/SKILL.md├── skills/ag-spec/SKILL.md├── skills/ag-version/SKILL.md└── skills/ag-wip/SKILL.md
Included Skills20
Abandon an Agentile spec that won't ship — record why, walk its dependent chain, and offer to cascade the abandonment so nothing is left silently blocked. Trigger phrases include "/ag-abandon", "abandon this spec", "drop this spec", "this didn't pass review", "kill this work item".
Take one Agentile spec from claim to shipped — claim the top prioritised ready spec (or a named one), plan it, implement it, verify it, ship it, and stop. Every pause is a checkpoint file in the spec's directory, so a factory worker and a person at a terminal run the same skill. Trigger phrases include "/ag-build", "build the next spec", "build <slug>", "work the next item".
Append a one-line stub to the Agentile inbox with today's date. Instant capture — no questions, no work, safe to run mid-build. Trigger phrases include "/ag-capture", "capture this idea", "drop a stub", "add to the inbox", "note this down for later".
Build out an Agentile stage playbook conversationally — interview the user about how a given loop stage should run in this project, then write/update .agentile/<stage>.md. Trigger phrases include "/ag-customise", "customise a stage", "configure the build/verify/... stage", "set up worktree-workflow for build".
Propose the inbox stubs that would make an Agentile Outcome true — four to eight one-liners the user accepts, edits or rejects, each landing in the Inbox linked to the Outcome. Never writes specs. Trigger phrases include "/ag-decompose", "decompose this outcome", "what work would achieve", "break down the outcome", "propose specs for".
Release the shipped specs sitting on trunk — run the project's pre-deploy checks and its deploy gate, then record what went out. Deploy is batched and separate from ship (which merges one spec). Trigger phrases include "/ag-deploy", "cut a release", "deploy to production", "release what's on trunk", "what's undeployed".
List the current Agentile inbox stubs and nothing else, so you can see what needs shaping at a glance. Read-only. Trigger phrases include "/ag-inbox", "show the inbox", "what's in the inbox", "what needs shaping", "list stubs".
Initialise Agentile in the current project — scaffold inbox.md, the project brief, the .agentile/ config layer, docs/adr/, and the CLAUDE.md standing-context section, and configure the deterministic gate commands. Idempotent; never overwrites existing files. Trigger phrases include "/ag-init", "set up Agentile here", "initialise Agentile", "scaffold the loop".
Retired alias for /ag-build (since 0.13.0). Runs /ag-build once and points at the replacement. Trigger phrases include "/ag-loop", "run the loop".
Show the Agentile world at Outcome level — each Outcome by rank with its specs by state, what is blocked, the free-standing specs, and the tag index. Read-only. Trigger phrases include "/ag-map", "show the map", "outcome view", "where are we", "what serves what", "the big picture".
Start a brand-new project on Agentile — interview the user for a project brief and stack, write the brief, a stack ADR and a starter CLAUDE.md, run the ag-init scaffold with gates pre-filled from the stack, then offer stage customisations and first inbox stubs. For fresh directories only; use /ag-init on an existing codebase. Trigger phrases include "/ag-new-project", "start a new project", "new Agentile project", "bootstrap a project with Agentile".
Pull the next piece of Agentile work — atomically claim the highest-priority unclaimed ready spec, stamp it with this session so it is resumable, and report it. Safe for concurrent loops. Trigger phrases include "/ag-next", "what's next", "pull the next item", "give me the next thing to work on".
Create or edit an Agentile Outcome — the falsifiable bet above specs — through a short interview for its claim, measure and stop rule; or record one as achieved. Non-coding. Trigger phrases include "/ag-outcome", "new outcome", "add an outcome", "edit the outcome", "mark the outcome achieved", "what are we betting on".
Turn a Ready spec into a written, reviewable plan before any code — promotes the spec to its directory form and writes plan.md (files to touch, approach, test strategy, risks) beside SPEC.md, using the ag-planner subagent or Plan Mode. Recommends an ADR for risky specs. Trigger phrases include "/ag-plan", "plan this spec", "plan the work", "propose an approach".
Interactively order the ready Agentile specs by assigning dense NNNN- filename prefixes that encode priority rank. Trigger phrases include "/ag-prioritise", "prioritise the backlog", "order the ready work", "re-rank specs".
Close the loop with a data-driven mini-retro — compile a flow digest from git history, specs, and ADRs (and PRs when available), surface where work waited and what needed rework, then propose concrete updates to CLAUDE.md and ADRs. Non-coding. Trigger phrases include "/ag-retro", "run a retro", "flow digest", "what did we learn", "weekly retro".
Promote a stub from the Agentile inbox to a Ready spec through a conversation — interview the user against the project's Definition of Ready, run the Business Value × Technical Certainty triage, then write the spec and remove the stub. Non-coding. Trigger phrases include "/ag-shape", "shape this stub", "shape an inbox item", "turn this into a spec", "make this Ready".
Write a Ready spec for a trivial change straight from a one-line idea, skipping the full shaping interview. Use only for small, certain work; non-trivial items should go through /ag-shape. Trigger phrases include "/ag-spec", "spec this", "quick spec", "write a spec for".
Report the installed Agentile version, and whether the running copy matches this repo. Read-only. Trigger phrases include "/ag-version", "what version of Agentile", "agentile version", "which Agentile is installed".
Show Agentile work in progress — which specs are claimed, by which session/label, since when, and how to resume or answer each. Read-only. Trigger phrases include "/ag-wip", "what's in progress", "what's being worked on", "show work in flight".
Plugin manifests1
{
"name": "agentile",
"version": "0.16.1",
"description": "Agentile for Claude — the Claude Code implementation of the Agentile methodology for 1–5 person teams who direct AI agents: capture → shape → spec → plan → build → verify → ship → learn. Ships the fixed loop machinery (skills, agents, hooks); each project tailors its content via scaffolded .agentile/ config.",
"author": {
"name": "Keith Rowell",
"email": "[email protected]"
},
"license": "MIT",
"keywords": [
"agentic",
"agile",
"methodology",
"workflow",
"spec-driven",
"lean"
],
"displayName": "Agentile",
"homepage": "https://github.com/agenta-consulting/agentile",
"repository": "https://github.com/agenta-consulting/agentile"
}For maintainers
If you maintain this plugin, link to this source-backed listing from your README so users can review its manifest and indexed components.
[agentile on Agent Plugins Marketplace](https://pluginsmp.com/plugins/agentile)