Guide · Agent skill
How to use the Request Progress agent skill (claude-mem)
Updated · written with AI from the project's own documentation and checked against our catalogue data
The request-progress skill is an agent skill for Claude that groups several separate conversations serving the same overall goal into one tracked request, then reviews each request using the agent's stored memory of past work. It is part of the claude-mem memory plugin and feeds the My Work dashboard. This guide explains what the skill does, how to install it, how to trigger it with natural requests, and what it needs to run.
Request Progressthedotmack/claude-mem · Apache-2.0 Download the skill Quick start
- 1Confirm requirements. Install the Bun runtime and the claude-mem plugin, and make sure its worker is running and its database exists at ~/.claude-mem/claude-mem.db.
- 2Download the skill ZIP. Download the request-progress ZIP from this page.
- 3Install the skill. Unzip it into ~/.claude/skills/ for personal use or .claude/skills/ inside a project, or upload the ZIP under Settings → Capabilities → Skills in the Claude apps.
- 4Ask for a grouping run. Say something like "group requests" or "group by overall request" to start grouping recent conversations.
- 5Review progress. Ask for "review request progress" or "refresh progress" to update each due request and its build story.
- 6Check My Work. Open My Work and check each card, since grouping can make mistakes.
What the request-progress skill makes the agent do
My Work shows one card per human request, built from the person's own prompts. People rarely finish a goal in one conversation: they clear the chat, switch worktrees, hand off to Codex, or come back the next day. This skill checks each new conversation's opening ask against the open requests in its project and joins related conversations into one overall request.
Once requests are grouped, the skill reviews each one from memory and writes a build story. The story's first paragraph becomes My Work's recap, and the story sheet shows shipping velocity, the narrative and learned working preferences. A review never changes a request's recorded status.
- Groups recent conversations into overall requests, within one project root
- Reviews each request's progress from claude-mem memory
- Sorts each request into one of your focus areas
- Writes the build story, including how released work landed
What you need before you install
The skill depends on claude-mem. Its scripts read claude-mem's local database and need its worker running, so the skill has no value on its own. The CLI is run with the Bun runtime.
An optional CMEM Pro or OpenRouter account enables an automatic judge that checks opening asks. Without it, conversations the judge cannot place are sent to the agent for grouping.
- Bun runtime, since the CLI is run with bun
- claude-mem plugin with its local database at ~/.claude-mem/claude-mem.db
- claude-mem worker running
- Optional: CMEM Pro or OpenRouter account for the judge's opening-ask checks
Install the skill from a ZIP or the Claude apps
Download the ZIP from this page and unzip it into your personal skills folder, ~/.claude/skills/, or into a project's .claude/skills/ folder if you want it only for that project. The skill's main file is SKILL.md.
Alternatively, upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside the project
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Trigger the skill with natural requests
You do not need a special command. Ask the agent in plain words. The skill's description lists phrases that should trigger it, so use wording close to these.
Default review window is 14 days. Ask for a different window and the agent will use it.
- "group requests"
- "group by overall request"
- "review request progress"
- "refresh progress"
- "build story"
- "shipping velocity"
- "how do I like to work with agents"
How grouping works and how to run it
Grouping decides once per conversation, from its opening substantive ask. A code link joins a conversation at once when it wrote to the same work-state list or recorded the same request_id as an open request. Requests stay inside one project root, so worktrees of one repo share requests.
The skill's CLI is run with a helper variable pointing at its scripts. The agent uses these commands, and you can run them yourself to inspect results. The judge command shows what it would do, and --apply writes the decisions.
- Preview judge decisions: judge --days 14 (add --apply to write them)
- Fetch an undecided batch: grouping-input --days 14
- Save decisions: apply, which reads JSON from standard input
- Check what is due for review: pending --days 14 --limit 48
- Show a request's details: show <groupId>
CLI="bun ${CLAUDE_SKILL_DIR}/scripts/request-groups.mjs"Safety notes: scripts and network
The skill runs scripts and uses the network. The CLI opens ~/.claude-mem/claude-mem.db read-only and writes only to ~/.claude-mem/work-view/. The one exception is record-focus, which saves through the running claude-mem worker because focus areas sync between computers.
The skill reads recent prompts and stored memory. If you do not want conversation content read or summarised, do not install it. Grouping can make mistakes, so check the cards in My Work. When the agent is unsure, it is told to keep conversations separate, because a wrongly merged card hides work.
- Reads your recent prompts and stored memory from the local database
- Writes only to ~/.claude-mem/work-view/ (except record-focus)
- Judge checks use an optional CMEM Pro or OpenRouter account
Tips for getting useful results
Review the cards after each run. Grouping is not guaranteed to be accurate, and the owner should confirm each card. If a group looks wrong, correct it rather than trusting the summary.
Keep requests in plain words. The skill's titles are meant to describe the goal, not the process.
- Check every card in My Work after grouping
- Use a window that matches the goal, such as the default 14 days
- Start with a small window if you have many conversations
What people use it for
- Tracking one feature built across several Claude Code sessions and a Codex handoff as a single card
- Getting a written build story of recent agent-assisted work for a team update
- Seeing progress meters and shipping velocity for a set of ongoing goals
- Reviewing learned working preferences by asking how you like to work with agents
Frequently asked questions
How do I install the request-progress skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
What does the request-progress skill require?
It requires the Bun runtime, the claude-mem plugin with its local database at ~/.claude-mem/claude-mem.db, and a running claude-mem worker. A CMEM Pro or OpenRouter account is optional and only used for the judge's opening-ask checks.
Is the request-progress skill free?
The source for this page does not state pricing or licence terms beyond the listed licence, Apache-2.0. The optional judge requires a CMEM Pro or OpenRouter account, so check those services for their own terms.
Can I use request-progress without claude-mem?
No. The scripts read claude-mem's local database and need its worker, so the skill has no value on its own.
Does the skill change the status of my requests?
No. A review never changes a request's recorded status. It updates progress, the build story and related details.
Will the skill group my conversations correctly every time?
Not guaranteed. The skill is told to keep conversations separate when unsure, but mistakes are possible, so check each card in My Work.
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
Use request-progress if you work on one goal across many Claude or Codex sessions, already run claude-mem, and want a My Work view that tracks requests rather than single chats. Pick something else if you do not run claude-mem or do not want your conversation content read and summarised.
Download the skill