In short:
ak-agentkitis a router: it picks one owning skill per intent from what is installed, chains them into a concise route, then hands the work off to that skill. When the next step is anakcommand, it classifies the command and settles project or global scope before running it. It does not do the real work itself, and an overwrite, delete or reset pauses for you when its exact effects are not yet authorized.
Disclosure: I maintain AgentKit and have an affiliate relationship with it: the AgentKit link in this post is an affiliate link; buying through it gets you 30% off.
Summary
Use it when you have multi-step work that spans several areas, or you do not know which skill owns it; or when the next step is installing, updating, checking, recovering or removing a kit with an ak command. You do not need it when you already know which skill owns the work: call that skill directly. Plain conversational questions do not need the router either.
flowchart LR
A["Multi-step work or unclear skill"] --> B["ak-agentkit"]
A2["Next step is an ak command"] --> B
B --> C["Hand straight to one skill"]
B --> D["Concise chain from installed skills"]
B --> E["Just a question: answer it"]
B --> G["Run the ak command in the right scope, report back"]
C --> F["Owning skill does the work, router stops"]
D --> FFour concepts to know:
| Concept | Question | Example |
|---|---|---|
| Proportionality gate | Does this work need routing at all? | You named a skill, or the work sits in one area with a clear owner: it hands off directly. A plain conversational question: it answers and runs no workflow. |
| Classification and the shortest chain | What kind of work is this, how big, how risky, and which steps are needed in which order? | "Add a profile settings page" is build-feature, size standard, risk elevated because users see it, 1 area. Work that touches money, security, data migrations or sending to many people is risk high. |
| Subagent handoff | Should this go to another agent, and how? | Three independent code areas to scout: hand them to 3 read-only agents running in parallel. Fixing one small file: do it directly, because handing it off costs more than the work. |
| ak commands: type, scope, source of truth | Will the ak command change anything, where, and do its flags really exist in the ak version you have installed? | ak doctor is read-only. ak plan kanban is diagnostic: it opens an interactive view for you to look at. ak update is a mutating command, so before running it the skill settles whether the change lands in the current project or in ~/.claude, ~/.codex, which affects every project. For flags, ak update --help or ak describe update --json beats the command table shipped with the skill; after ak self-update, the binary can be newer than that table. |
Sample prompt for work that spans several areas:
/ak-agentkit Add plan-based billing with Stripe and a plan settings page for the <app name> app.
Outcome: users pick a plan, pay, and see the new plan on the settings page.
Constraint: do not change the existing sign-in flow.
Done when: billing tests pass in test mode, with an independent review before shipping.This skill has no flags.
Below is the detailed version: the run flow, 6 situations with prompts, the change history by ak release, and common pitfalls. The detailed version and the interactive page are for subscribers. Your agent can also search them through the portal's MCP.
Opens when your account has subscriber access
This guide is for subscribers
Once your account has access, the full piece, the interactive visual and the agent call all open.
- Sign in or create a modernEngineer accountReading access belongs to your account email.
- Choose a plan on the pricing page1, 3 or 12 months, paid by VietQR bank transfer. Access opens about a minute after the money arrives.
- Or redeem a Facebook subscriber codePeople supporting the page on Facebook get an activation code by message each month.
Already a subscriber? Sign in to read.
Use this piece in your agent
Get an API keyCopy the text below and paste it into Claude Code, Codex or whichever agent you use. The agent reads the piece and checks it against the repo you have open.
Read "ak-agentkit: pick skills, chain the work, run ak commands safely" on modernEngineer and apply it to this repo.
1. Call the content_get tool of the modernengineer MCP server with locale "en", slug "ak-agentkit". If the modernengineer MCP server is not connected, stop and walk me through connecting it from https://modernengineer.dev/en/mcp.
2. Summarize the main steps of the piece and point out which apply to this repo.
3. Propose concrete changes and ask me before editing any file.Connect MCP in Claude Code (once)
claude mcp add --transport http modernengineer \
https://api.modernengineer.dev/mcp \
--header "Authorization: Bearer <API key>"Reading the full piece over MCP needs an API key from a subscriber account or one that bought this guide. Without access, the agent only gets the opening.
- Version
- v1.1.0
- Updated
- 01/11/2026
- Licence
- All rights reserved
- Slug
- ak-agentkit
