How to use the Owner Goals Mapper agent skill

What does the Owner Goals Mapper skill make the agent do?

When you ask, the agent reads your past Claude Code and Codex prompts and leaves out any prompt an agent wrote for itself. It then sorts each statement you made into exactly one of three kinds: a focus area, which is a part of your work you can ask for things inside of; an ask, which is one thing you requested that can be marked done; or a rule, which describes how you want work done.

Each item keeps your own words and the date and time you typed them. The results are saved to claude-mem's My Work view: focus areas appear in My Focus as suggestions and rules appear in My Rules as suggestions. You can keep, rename, merge or dismiss each one.

How do you install the skill?

Download the ZIP file from this page. Then choose one of two routes.

For a single user on your own machine, unzip the file into ~/.claude/skills/. To limit the skill to one project, unzip it into that project's .claude/skills/ folder instead. In the Claude apps you can also upload the ZIP under Settings → Capabilities → Skills.

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

How do you trigger the skill with a natural request?

You do not need a command. Ask in ordinary language. The skill's description lists the phrases that should start it, including 'owner goals', 'my goals', 'what do I want', 'what am I asking for', 'map my prompts', and requests to refresh Goals in My Work.

Asking for a specific period also works. The skill can map a different number of days or a fixed range such as a past month.

• Show me my owner goals for the past month.
• What am I asking for across my sessions this week?
• Map my prompts and save the result to My Work.
• Refresh my goals in My Work without mapping everything again.

What happens when the skill runs?

The skill works in steps. First, an export script collects your prompts from local transcript folders into a working folder inside the current project. The script reports how many prompts it kept and how many it excluded. Second, the agent saves the focus areas already in My Focus so it can tell which new areas match them. Third, a subagent reads the exported prompts and writes a map with a summary, what you want, how you want work done, what frustrates you, open asks, and ranked focus areas. Every point quotes your exact words with a timestamp.

The subagent then sorts the map into focus areas, asks and rules and saves them to My Work.

• Export: reads ~/.claude/projects and ~/.codex/sessions, then writes owner-prompts.md and excluded-agent-written-prompts.md
• Map: writes owner-map.md with quotes and timestamps
• Save: writes owner-goals.json and sends the focus areas, asks and rules to My Work

How do you update saved goals without mapping everything again?

If goals were saved before, the skill can update them instead of starting over. It exports only the prompts typed since the saved generated_at time, into new-owner-prompts.md. It then puts those new statements through the same tests and updates the saved goals. Titles stay the same for items that did not change, so requests already linked to a focus area stay linked. If there are no new prompts, the skill saves the existing goals again with a new timestamp.

What does the skill need before it runs?

The skill needs Python 3.9 or newer. It uses only the Python standard library, so there are no packages to install. It also needs local Claude Code transcripts in ~/.claude/projects, Codex sessions in ~/.codex/sessions, or both. To receive the saved goals, claude-mem must be installed with a running My Work view.

On Windows, the source replaces python3 with python -X utf8, or py -3 -X utf8 when using the Python launcher. The Microsoft Store python3 alias is not a real interpreter.

• Python 3.9 or newer
• Local Claude Code transcripts and/or Codex sessions
• claude-mem with a running My Work view

What should you check for privacy and safety?

The skill runs scripts on your machine and reads local transcript folders. It does not use the network, according to the skill's listing. The scripts store quotes of your prompts in the working folder, and those quotes can include private or confidential text you typed. The export script adds the working folder to the checkout's local git exclude file when git would otherwise track it.

Do not run this skill on a shared or sensitive machine unless you are comfortable with that stored text. Review owner-prompts.md before you share any folder from the project.

• Runs scripts: yes, Python scripts in the skill's scripts folder
• Uses network: no
• Stored output includes verbatim quotes of your prompts
• The working folder is added to local git exclude when git would otherwise track it

What are the limits and useful tips?

The map is an interpretation of your prompts, not a complete task list. It can miss items or misjudge them, so treat it as a starting point. Use it to spot recurring goals and frustrations over a period, not to track every task exactly.

Keep the time window in mind. A 31-day default captures a month of prompts. For a narrower question, pass a smaller window or a specific date range. Review suggestions in My Work before you keep them, since you can rename, merge or dismiss anything.

• Review each suggestion before keeping it
• Use a specific date range when you want a fixed period
• Use the update route for regular refreshes