Logbooks are shared working state: structured entries that agents and humans can append, patch, annotate, query, and carry forward across sessions. Learn more at agentlogbooks.com.
This repo is a collection of Claude skills that use logbooks as their working surface.
Works in both Chat and Cowork modes — the plugin installs once and is available everywhere.
In Chat mode:
- Click Customize in the left sidebar
- Open the Directory and select the Plugins tab
- Switch to the Personal tab, click "+" → Add marketplace
- Enter
agentlogbooks/logbooksand click Sync
In Cowork mode:
- Click Customize in the left sidebar
- Next to Personal plugins, click "+" → Add marketplace
- Enter
agentlogbooks/logbooksand click Sync
Upload manually:
- Go to github.com/agentlogbooks/logbooks → green Code button → Download ZIP
- Rename the file from
.zipto.plugin - In the same plugin menu above, choose Upload plugin instead and select the file
/plugin marketplace add agentlogbooks/logbooks
/plugin install logbook-creator@logbooks
/plugin install deep-code-review@logbooks
/plugin install ideation@logbooksOr from a local clone:
git clone https://github.com/agentlogbooks/logbooks.git ~/logbooks
claude --plugin-dir ~/logbooks/plugins/logbook-creator| Plugin | Skill | What it does |
|---|---|---|
logbook-creator |
/logbook-creator |
Design and create a logbook — a shared, queryable, schema-stable working surface for agents and humans to append, annotate, and query across sessions. |
deep-code-review |
/deep-code-review |
Hotspot-first deep code review: change map → risky hotspot selection → per-hotspot lens subagents → skeptic pass → comment budget. Surfaces at most 5 high-signal findings or questions per run. |
ideation |
/ideation |
Idea-centric brainstorming on a per-topic SQLite logbook. Atomic operators (generate, transform, evaluate, validate, decide) compose into playbooks. Covers brainstorming, naming, developing or combining ideas, stress-testing candidates, reframing, and picking final directions. Ideas persist across sessions — follow-up prompts resume where the last one ended. |
Ask Claude to create a logbook when you need to track structured entries across sessions, stage drafts before pushing to a target system, or coordinate state between agents. Invoke the skill directly with:
/logbook-creator
Claude will walk you through the motivation, schema, and storage choice, then write the logbook file and a sibling spec describing it.
Run a hotspot-first code review on a PR, branch, or diff:
/deep-code-review review PR #42
/deep-code-review review current branch changes
The skill maps behavior changes, selects risky hotspots, runs per-hotspot analysis, applies a skeptic pass, and surfaces at most 5 high-signal findings or questions. Results are persisted to a per-PR JSONL trace and SQLite ledger at ~/logbooks/code-review/.
Start or continue ideation on a topic:
/ideation coffee-shop-q2: brainstorm ways to grow afternoon revenue
/ideation coffee-shop-q2: develop idea 17 further
/ideation product: what should we call this product
/ideation shortlist --playbook stress_test_shortlist: validate the top 3
Default for a fresh topic is a lightweight flow (frame + ~20 ideas + a compare report). Opt in to deeper shapes with explicit intent — "thorough" / "deep" for the full treatment, "score them" / "prioritize" for a scored pass, "name X" for the naming playbook. Logbooks live at ./.ideation/<slug>/logbook.sqlite and persist across sessions, so follow-up prompts on the same slug pick up where the last one ended.