Generates on-demand summaries of your current session
Generates on-demand summaries of your current session, active workflow, or entire project history. Pulls data from Cowork’s file-based sources — git log, chat history, recipe audit log, and MPCS/CPM state files — and formats it at your chosen detail level.
Cowork Session Summary
Tags: Introduced in Beta, Summary, Session, Workflow, Project, Status, Git, Bridge, Cowork
TL;DR
What It Does
Generates on-demand summaries of your current session, active workflow, or entire project history. Pulls data from Cowork’s file-based sources — git log, chat history, recipe audit log, and MPCS/CPM state files — and formats it at your chosen detail level.
How It Works
Two parameters drive the output. Verbosity controls detail level: brief gives a 2–3 sentence snapshot, medium gives a structured report with accomplishments and git status, and detailed gives the full picture with files, decisions, and recipe audit. Scope controls what data is gathered: session reads the current session’s git log, workflow reads .mpcs-state.json or .cpmrc.json for active multi-persona state, and project reads the chat history file for cross-session cumulative accomplishments.
What To Expect
A formatted status message displayed in the conversation. No files are written — the summary is ephemeral. The level of detail scales with your verbosity choice, from a quick one-liner to a comprehensive session report with git diffs and recipe execution counts.
Best Results When You
Ask for a summary at natural checkpoints — after completing a task, before a handoff, or when you need to orient yourself mid-session. Use scope=workflow when you have an active MPCS or CPM engagement. Use scope=project for cross-session overviews when you need the big picture.
Typical Time
Under 1 minute for all verbosity levels. The recipe reads existing files rather than generating new content.
Difficulty
Beginner — just ask for a summary. The recipe auto-detects verbosity and scope from natural language, so “quick summary” maps to brief/session and “full project overview” maps to detailed/project without needing explicit parameter syntax.
How To Start
This recipe provides session, workflow, or project summaries at three detail levels. Most users simply ask for a summary and the recipe auto-detects the right parameters.
STEP 1Request Summary
Ask for a summary in natural language. The recipe detects your intent and selects the right verbosity and scope.
Available parameters
-
verbosity
· string · optional · default: medium
Summary detail level: “brief” (2–3 sentences), “medium” (structured report), or “detailed” (full report with tables and audit trail). Auto-detected from natural language. -
scope
· string · optional · default: session
What data to include: “session” (current session only), “workflow” (MPCS or CPM state), or “project” (cross-session history from chat history file).
Invocations
Quick session snapshot
Give me a brief summary
Standard session summary
Summarize this session
Full project overview
Give me a detailed project overview
Usage Examples
EXAMPLE 1Quick mid-session check
User
Brief summary.
Cat
This session: 4 tasks done, 2 pending. Currently writing recipe page content. Next: CWK-035 pipeline.
EXAMPLE 2Standard session summary
User
Summarize this session.
Cat
Session Summary. Accomplishments: Completed CWK-032 through CWK-034 pipeline runs. All three BLESSED. Files created: 15 (5 per recipe). Current State: Processing forward through COWORK cookbook. Git Status: main @ a3f7c2e | dirty with 15 changes. Next Steps: Continue with CWK-035, target 6 recipes this session.
EXAMPLE 3Workflow status check
User
Show workflow status.
Cat
MPCS Workflow Status. Project: Deployment Guide. Content: API Guide (documentation). Active Personas: Auguste (Creator, E), Fernand (Validator, E). Cycles: 2. Latest version: v1.1. Status: ROUTED. 1 pending question for Creator.
EXAMPLE 4Full project overview
User
Detailed project overview.
Cat
Detailed Session Summary. Session: H042 | Date: 2026-04-20. Scope: project | Persona: Cat (E). Accomplishments: 42 sessions, 128 recipe pages processed, SP01–SP09 completed. Files Modified: [table]. Key Decisions: [list]. Git Log: [last 10 commits]. Recipe Audit: [execution counts]. Deferred Items: Phase 3 LOCK amendment, GitHub permalinks. Next Steps: Continue forward processing.
How AI Reads This Recipe
When this recipe is triggered, the AI generates a formatted summary. The AI should:
- Detect Parameters (Step 1). Parse the user’s request for verbosity cues (“brief”/“quick” → brief, “detailed”/“full” → detailed, else → medium) and scope cues (“workflow”/“mpcs”/“cpm” → workflow, “project”/“all sessions” → project, else → session).
- Gather Data (Step 2). Always run git log (last 10 commits) and git status. Use session context if available. For scope=project, read the chat history file and extract accomplishments and key decisions. For scope=workflow, read .mpcs-state.json and/or .cpmrc.json. If a recipe audit log is available, extract this session’s recipe executions.
- Check Minimal Activity (Step 3). If no tasks completed and no files created in a session-scope summary, return an early note rather than an empty summary.
- Generate Summary (Step 4). Format output according to verbosity level. Brief: one sentence with counts and next step. Medium: structured blocks for accomplishments, current state, git status, and next steps. Detailed: full report with session metadata, file modification table, key decisions, git log, recipe audit trail, deferred items, and prioritized next steps.
- Scope-Specific Handling (Step 5). For workflow scope: parse and display MPCS state (active personas, routing, phase) and/or CPM state (current phase, completed phases, blockers). For project scope: display cross-session cumulative summary with handoff count, date range, and total accomplishments. Caveat accuracy depends on chat history completeness.
- Deliver (Step 6). Display formatted output. No file write — the summary is ephemeral. Log an audit trail entry with recipe ID, parameters, and data sources used.
Error handling: missing chat history → fallback to session scope. Corrupted workflow state → suggest CWK-031 INSPECT. Empty git log → session state only. Unrecognized verbosity or scope → use defaults.
When To Use This Recipe
Use this recipe when you want a quick status check at any point during a session, need to orient yourself after returning from a break or context switch, want to verify what has been accomplished before writing a handoff, need an MPCS or CPM workflow status check, or want a cross-session project overview from chat history.
Do Not Use When
You need a persistent written summary — this recipe produces ephemeral display output only. For a written handoff, use CWK-002 (Session Handoff). For a written project report, use CWK-032 as an input to inform what you write, but the writing itself is a separate task.
Recipe FAQ
Does this recipe write any files?
No. The summary is displayed in the conversation only. It is ephemeral output. If you need a persistent record, use CWK-002 for session handoffs.
How does auto-detection work?
The recipe scans your request for keywords. “Brief” or “quick” sets verbosity to brief. “Detailed” or “full” sets it to detailed. “Workflow”, “mpcs”, or “cpm” sets scope to workflow. “Project” or “all sessions” sets scope to project. Anything else defaults to medium verbosity, session scope.
What if the chat history file is missing?
If scope is project and the chat history file cannot be found, the recipe warns you and falls back to session scope. You still get a summary, just limited to the current session’s data.
What data sources does this recipe use?
Git log and git status (always), session context (if available), chat history file (project scope), .mpcs-state.json and .cpmrc.json (workflow scope), and recipe audit log (if available).
What is the difference between session and project scope?
Session scope summarizes the current session only — what you have done since the session started. Project scope reads the chat history file to summarize across all sessions, giving a cumulative view of accomplishments, decisions, and progress.
Version History
v1.00a — 2026-03-02
Initial creation. Adapted from STU-032 for Cowork’s stateless file-based architecture. Three verbosity levels, three scope options, auto-detection from natural language, git integration, MPCS/CPM state file parsing, and recipe audit log extraction.
Get this recipe with CRAFT for Claude Cowork
Cowork recipes ship bundled with CRAFT for Claude Cowork — there’s no separate download. Clone the framework once, and your AI runs every recipe automatically when invoked.
Pull anytime to stay on the latest version — free to clone, no login or email required.
Then start your session
Once CRAFT is in your project folder, open a new Cowork session and ask Claude to initialize. For example:
You
Please initialize my CRAFT session.
Claude
CRAFT session ready. Your project is loaded, your persona is active, and your recipes are available. What would you like to work on?
Not familiar with Git? Download as a ZIP
No command line needed. Just download, move, and unzip:
- Open the CRAFT framework repo on GitHub.
- Click the green Code button, then choose Download ZIP.
- Move the downloaded ZIP into your Claude Cowork project folder.
- Unzip it: double-click on Mac, or right-click → Extract All on Windows.
