Suggestions come to you automatically, but a wait is not necessary. A set of /tacit: commands lets you pull from the registry at any time: before a task, during a task, or after a task.

The first three commands (search, review, and audit) do the same playbook retrieval on three different inputs: a query that you give, your current work, and a completed session. Use the command that matches your scope. The evidence is the same for all three.

Because these commands are pulls, they do not change the funnel. This type of playbook access records no shown event, so it does not increase the adoption or helped rate of a technique. (A pushed ◆ suggestion and a technique that an @tacit answer cites do count, because a member saw the technique.)

Search the playbook

/tacit:search deploying a lambda behind an api gateway

OpenTacit searches the playbook of your organization and reports the best matches (a maximum of four). Each match includes:

• the name and the id
• the measured evidence, quoted without a change
• one sentence that tells you when the technique applies
• the recipe, ready for use.

If you run the command with no query, OpenTacit makes a query from your current work.

A result with no matches is a real answer: your organization has no validated move for that subject. When you find a move that works, contribute one.

Review your current work

/tacit:review

/tacit:review examines your current work: the full session, or a focus that you give. It finds the different activities in the work and checks each activity against the registry. Matches show as ◆ OpenTacit blocks in the conversation. Because of this, the command operates on all surfaces. This includes Claude Code on the web and on phones, where automatic suggestions cannot appear.

You can also ask in your own words ("review this against OpenTacit"). This runs the same check in a subagent that examines more deeply.

Audit a finished session

/tacit:audit

/tacit:review examines current work. /tacit:audit examines a completed session: the current session, a transcript file, or a shared conversation link. The audit shows the techniques that the registry had for that session. It then proposes follow-up actions. Each action needs your confirmation:

• A technique to try the next time
• A repeated problem that the registry does not cover, with an invitation to contribute
• A revision for a technique that gave incorrect guidance; the revision goes for review, and the serving technique does not change

You can run the same audit from the command line: tacit audit <file-or-url>.

Ask for usage data

Three commands answer the larger questions from your session:

• /tacit:insights [7d|30d|90d|all] — the pipeline of technique usage: the shown → adopted → helped funnel, the helped rate, and trends. In clients with inline panels, the output shows as a small dashboard. In the terminal, it shows as text.
• /tacit:org [7d|30d|90d|all] — the view of who and where: the techniques that each cohort adopts, the areas of practice that grow, and the places where the proven move of one team can help another team. The data is always cohort aggregates, never individuals.
• The playbook map — in clients with inline panels, ask for "the tacit map" to see the full live playbook as a graph. The position of each technique shows its relation to the other techniques. The size shows adoption. The shade shows the helped rate.

The dashboard gives all of this and more in the browser. This includes the Events feed, a continuous record of OpenTacit activity.

Review drafts without the dashboard

/tacit:drafts

This command runs the draft review queue in your session. You make the same decisions as on the Review page of the dashboard, but you stay in your work. The command uses the best mechanism that your client supports:

• An interactive panel — in clients that show MCP apps (Claude on the web and desktop), the queue shows the full technique of each draft, with Promote and Reject buttons. Decisions apply immediately.
• A native form — in the terminal, OpenTacit shows each draft in full. A one-keystroke question asks for your decision: promote, reject, edit first, or skip.
• Plain conversation — on all other clients, OpenTacit shows the drafts in the chat. OpenTacit does an action only when you give an instruction.

With each mechanism, OpenTacit promotes or rejects a draft only when you give an explicit instruction. If your member key cannot decide drafts, the command tells you this and refers you to the dashboard.