How to use the Weekly Digests agent skill (weekly-digests)

What is an agent skill and what does weekly-digests do?

An agent skill is a folder containing a SKILL.md file that gives an AI agent instructions and helper scripts for one specific task. The agent reads the file when it decides the task applies to your request.

The weekly-digests skill (from thedotmack/claude-mem, Apache-2.0) works on a project's full claude-mem timeline. It splits the timeline into per-ISO-week files, then writes one digest chapter per week in order. Each chapter's writer receives a carry-forward block summarising the prior week so the narrative stays continuous. The output is a folder of digest files plus a weekly index table.

• The chapter count equals the number of ISO weeks the timeline covers.
• Each weekly source file keeps the original daily sections verbatim before the digests are written.
• Use it when you want a readable, chronological story of how a project evolved.

How do you install the weekly-digests skill?

Download the ZIP from this page and unzip it. Then place the resulting skill folder in one of two locations: your personal skills folder at ~/.claude/skills/, or a project-specific folder at .claude/skills/ inside the project you are working on.

Alternatively, upload the ZIP in the Claude apps under Settings → Capabilities → Skills.

• Personal install: ~/.claude/skills/ (available across your projects).
• Project install: .claude/skills/ inside the project folder (available only there).
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP.

What do you need before running it?

The skill depends on a local claude-mem worker and several standard tools. Check these before you ask for a digest, because the pipeline is multi-step and will fail partway if something is missing.

• claude-mem worker running locally.
• Node.js, used to read the worker port settings.
• Python, for the timeline splitting script.
• curl and git available in the shell.
• A writable output directory such as docs/timeline-weeks.
• The project must have at least one ISO week of observations.

How do you trigger the skill with a natural request?

You do not need special syntax. Ask in plain language, using phrases the skill is written to respond to. The skill is triggered by requests such as the ones below.

• "Weekly digests" of this project.
• "Week-by-week story" of the project's history.
• "Serial timeline" of the project.
• "Narrative chapters" of the project's history.
• "Continue the story week by week."
• "Run a digest for each week."

What happens when the skill runs?

The agent first determines the project name, handling git worktrees by using the parent project. It then fetches the full timeline from the local worker and saves it to .scratch/cm-timeline.md. A Python script writes .scratch/split-timeline.py to split the timeline into one file per ISO week under docs/timeline-weeks/, and a README.md index table is created.

Subagents then write the chapters one at a time. Each one runs only after the previous agent finishes, because each needs the prior week's carry-forward block. Each digest is saved in docs/timeline-weeks/digests/.

• Output: one digest file per ISO week plus a weekly index table.
• Each digest includes a title line, tagline, narrative, threads, a cliffhanger and a carry-forward block.
• The process is slow and multi-step, so expect a long run for long histories.

What are the safety notes for running weekly-digests?

This skill runs scripts and uses the network. It fetches data from a local worker URL on localhost, writes Python and Node-based helper commands, and creates folders and files in your project. Read the skill before running it if you are unsure what it will execute.

The timeline may include security-related or sensitive observations, and these can be copied into the written digest files. Think about whether those files will be committed or shared. Only run it where the agent is allowed to write into the output folders and launch many subagent runs.

• Runs scripts: yes. Uses network: yes (localhost worker).
• Sensitive observations (security alerts, notes, or items marked sensitive) may appear in output files.
• Choose a clean output directory you are comfortable writing into.
• Do not use it where the agent should not write to arbitrary folders or launch many subagent runs.

Tips for better digests

Count the weeks before you start so you know how many chapters to expect. A history with many weeks produces many files and a long run. For a single short overview, or for a single long report, a different approach fits better; the timeline-report skill is the documented alternative for long histories.

Review the weekly index table in docs/timeline-weeks/README.md. It shows observation and session counts per week, which helps you see peak and quiet periods.

• Use a clean output folder such as docs/timeline-weeks.
• Check the sensitive content of the timeline before sharing the digests.
• Expect quiet weeks to produce short chapters; the skill is written to avoid padding them.