How to use the Skill Router Map agent skill (map)

What the map skill does for your agent

The map is a set of tables listing which skill owns which job, arranged in delivery-loop order: external work arriving, requirements, planning, execution, review, verification, pull requests and environment learning. It also lists on-ramps such as debugging, tests, and memory, plus utilities such as setup, diagnostics and cancellation.

The map does not do the work itself. It only says which skill to use, and in what order when more than one fits. The routed skills are separate files that are not included with the map, so they must be installed separately.

• Ordinary rows resolve to one Skill-tool call, after which the agent follows that skill's own instructions.
• Some routes are marked manual; for those the map shows you a slash command to run yourself, because model invocation is disabled for this skill.
• If two rows both fit, the earlier step in the delivery loop wins.
• If no row matches a job, the map says so plainly instead of guessing the nearest skill.

Install the map skill

Download the ZIP for this skill from our page. Then choose one of the two install routes below, depending on where you want the skill available.

For Claude Code, unzip the file into ~/.claude/skills/ to make it available in all your projects, or into a project's .claude/skills/ folder to limit it to that project. In the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Install the oh-my-claudecode skill set too

The map only points to other skills. Those skills, such as plan, execute, review and verify, are defined elsewhere in the oh-my-claudecode project and must be installed for the routes to resolve. Without them, the map will point at skill names that do nothing.

The map skill itself needs Claude Code or a compatible agent that supports the Skill tool.

How to trigger the map with natural requests

You do not need a special command. Describe the job you need done and say you are not sure which skill fits. The skill's description tells the agent to use it when something needs doing but it is not clear which shipped skill owns the job.

The skill also accepts an argument hint, "[what are you trying to do]", so you can state the goal directly.

• "I have a vague idea for a feature. Which skill should I start with?"
• "Someone sent me a chat log as requirements. What is the next step?"
• "A bug is showing up in production. Which skill handles root cause analysis?"
• "I think this is finished. Which skill checks that it really works?"
• "Work is ready to go out as a pull request. Which skill do I use?"

Examples of routing decisions

Each example shows how the map's tables lead to a skill. Read the routes as a guide to the choice, not a guarantee of the exact skill the agent will pick, because the model reads the tables and can vary its choice.

• Vague idea with unclear scope: the table points to deep-interview, which asks questions before anything else.
• Approved plan ready to build end to end: the table points to launch.
• A bug needing its root cause: the on-ramp table points to debug, with trace for competing-hypothesis tracing.
• Session that has pivoted several times and needs reorientation: the map says to run /oh-my-claudecode:session-flow manually.

What the skill needs and safety notes

The map skill runs no scripts and uses no network access, according to its source. It is a plain instruction file (SKILL.md). Its actual behaviour depends on the routed skills being installed, and those are not included in this download.

Because routing is decided by the model reading the tables, the same request may be routed differently on different runs. If you need a fixed, auditable process, do not rely on the map alone.

• Runs scripts: no
• Uses network: no
• Required: the oh-my-claudecode skill set and Claude Code or a compatible agent with Skill tool support

Tips for getting good routes

Describe the outcome you want rather than the tool you think you need. The map is organised by job, so a clear job description matches the right row more reliably.

If the agent reports that a job matches no row, treat that as useful information. The map is designed to say so rather than pick a near-miss skill.

• State what has happened so far, such as whether requirements are clear or a plan is approved.
• Ask for the slash command when a route is marked manual.
• Check that the target skill is installed if the agent reports it cannot find one.