Guide · Agent skill
How to use the Desk Journal agent skill (desk-journal)
Updated · written with AI from the project's own documentation and checked against our catalogue data
Desk Journal is an agent skill that teaches an AI agent to keep a running journal file for a named work area, called a desk. It is useful when work on a project spans many separate sessions, because the agent writes a short entry at the end of each session and reads the latest one at the start of the next. This guide explains what the skill does, how to install it, how to trigger it, and what to watch for.
Desk Journalgithub/awesome-copilot · MIT Download the skill Quick start
- 1Download the ZIP. Get the desk-journal ZIP from the item page on this site.
- 2Unzip the skill folder. Unzip it into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project. Make sure SKILL.md is in the skill folder.
- 3Or upload it in the Claude apps. In the Claude apps, go to Settings, then Capabilities, then Skills, and upload the ZIP.
- 4Start a long-running desk. Ask the agent to work on a named desk, so that a desks/<desk-name>/journal.md file can be created and used.
- 5Write an entry at session end. Ask the agent to write a journal entry with what was done, current state and next step, and to include dead ends.
- 6Resume in the next session. At the start of the next session, ask the agent to read the desk's journal and pick up from the most recent entry.
What is an agent skill, and what does Desk Journal do?
An agent skill is a folder of instructions that an AI agent can load and follow when a task matches it. The skill is defined by a file named SKILL.md. Desk Journal is one such skill, named desk-journal, and it is licensed under MIT.
Desk Journal is about memory between sessions. The skill describes the journal as the persistent memory that survives session boundaries. The agent appends entries to a file, and a later session can read them to pick up where the work stopped without rediscovering everything.
- End of session: write what was done, the current state and the next step.
- Start of session: read the journal and resume from the most recent entry.
- Mid-session checkpoint: note significant progress or decisions.
- Desk wind-down: write a final summary when the desk is being closed.
Install the Desk Journal skill
Download the ZIP for the desk-journal skill from the item page on this site. Unzip it so that the skill folder, containing SKILL.md, sits in one of these locations.
For a personal setup available across projects, use ~/.claude/skills/. For a single project, use the .claude/skills/ folder inside that project.
If you use the Claude apps instead, you can upload the skill under Settings, then Capabilities, then Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into .claude/skills/ inside your project
- Claude apps: Settings, then Capabilities, then Skills, then upload the ZIP
How to trigger Desk Journal with natural requests
You do not need a special command. Ask the agent in plain language, and make clear whether you want it to write, append or read a journal entry for a named desk. The skill covers writing, appending and reading, so describe which one you need.
Name the desk when you can. The journal path in the skill is built from the desk name, so a clear name helps the agent find the right file.
- Wrap up this session and write a journal entry for the migration desk.
- Read the journal for the security-audit desk and tell me where we left off.
- Add a mid-session checkpoint to the reporting desk journal about the schema decision.
- We are closing the onboarding desk. Write the final desk summary.
What a good journal entry looks like
The skill writes each entry to desks/<desk-name>/journal.md, appended as a section. Each entry has a date and short summary in the heading, followed by three fields: what was done, the current state, and the next step.
The skill asks for specific entries. Vague notes such as worked on security scanning do not help the next session. Entries should also keep to about three to five lines, and should record dead ends so they are not repeated.
- Be specific about what was done and what the results were.
- Record what did not work, so the next session does not repeat it.
- Keep it short: three to five lines per entry.
- Always include a next step, because the next session starts from zero.
## <date> — <short summary>
- **Worked on:** <what was done this session>
- **Current state:** <where things stand right now>
- **Next step:** <what the next session should pick up>Close a desk with a final entry
When a desk is being wound down, not just a session ending, the skill uses a separate closing entry. It records what the desk accomplished overall, what is on the bench from that desk, and anything another desk or the operator needs to know.
The skill says that if more context is needed than fits in a short entry, that context should be saved on the bench as a separate artifact rather than added to the journal.
- Summary: what the desk accomplished overall.
- Artifacts: what is on the bench from this desk.
- Handoff: anything another desk or the operator needs to know.
## <date> — Desk closed
- **Summary:** <what this desk accomplished overall>
- **Artifacts:** <what's on the bench from this desk>
- **Handoff:** <anything another desk or the operator needs to know>Safety notes before you use it
Desk Journal runs no scripts and makes no network calls, according to its listing. It requires nothing beyond the skill file itself.
Journal entries are plain text files that persist between sessions. The skill does not secure them. Do not use the journal for sensitive data, such as credentials or private personal information, and keep the desk folder somewhere you are comfortable storing plain text.
- Runs scripts: no
- Uses network: no
- Requirements listed: none
- Entries are plain text and are not secured by the skill
What people use it for
- Multi-week refactors where each session starts by reading where the last one stopped.
- Teams handing work between people or agents through a readable trail in the journal file.
- Investigations where recording dead ends stops the agent repeating failed approaches.
- Winding down a finished workstream with a closing summary that lists its artifacts and handoff notes.
Frequently asked questions
Is the Desk Journal agent skill free?
The skill is listed under the MIT licence on the source repository. Check the licence file in the source for the full terms before reuse.
What do I need to run the desk-journal skill?
The listing says it requires nothing beyond the skill itself. It runs no scripts and makes no network calls, so there are no extra dependencies listed.
Where do I put the desk-journal skill folder?
Unzip it into ~/.claude/skills/ for personal use or into .claude/skills/ inside a project. In the Claude apps, you can upload the ZIP under Settings, then Capabilities, then Skills.
Where does the agent save journal entries?
The skill appends entries to desks/<desk-name>/journal.md, where the desk name is the one you give the work area.
Is Desk Journal safe to use with confidential data?
Not for sensitive data. Entries are plain text files that persist between sessions, and the skill does not secure them, so keep secrets and private information out of the journal.
When should I not use Desk Journal?
Skip it for one-off questions or short tasks that finish in a single session, since a journal adds little value there. Use it for long-running projects that span many sessions.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Desk Journal suits long-running projects where work spans many agent sessions, and teams that want a readable handoff trail or a record of dead ends. It is a poor fit for one-off or single-session tasks, and for any work where the journal would hold sensitive data, because the entries are unsecured plain text.
Download the skill