clickmax-playbooks
Use when the user wants to create, edit, delete, link to pipelines, preview, or rehearse a CRM playbook (the sales method with stages, questions, tones, keyword triggers and guardrails that guides meetings and calls) in Clickmax.
How do I install this agent skill?
npx skills add https://docs.clickmax.io --skill clickmax-playbooksIs this agent skill safe to install?
No partner audit is available yet. Read the source before installing.
What does this agent skill do?
As tools abaixo aparecem com os nomes que o MCP da Clickmax registra. Se o seu cliente de IA prefixar nomes de tool (
mcp__<servidor>__,mcp_<servidor>_, ou outro), use o nome já prefixado que aparecer na sua lista de tools.
When this applies
Use for the sales-method document in Contatos & Comercial › Configurações › Playbooks (Contacts & Sales › Settings › Playbooks): write or change one, make pipelines suggest it, preview what a meeting room receives, rehearse it with the AI copilot, or delete it.
Not this skill:
- pipeline board, stages, cards, pipeline settings ->
clickmax-pipelines - lead search / identity ->
clickmax-leads - an AI agent's own persona/knowledge -> not exposed here; this skill only picks an existing agent by id
- what an AI SDR agent asks or how it qualifies -> NOT a playbook: the agent never reads one (Settings › Meus agentes, persona + Flows block; no tool here edits it)
Key assumptions
- A playbook is an independent entity, not a pipeline setting. Pipelines only SUGGEST one when a meeting is booked; whoever books can pick another playbook or none. Suggestion applies only when the booking does not choose (public booking pages, automations, AI bookings, or a
scheduled_calls_createthat omitsplaybookId). - Effect follows the
scopes:meetingis the Click Meeting room panel (Roteiro column + AI copilot briefing) and the voice-call window;livechatis the playbook block of the LiveChat lead cockpit (progress kept per conversation,playbooks_listwithscope: "livechat"lists the eligible ones).whatsappandflowsare shown as "coming soon" and are rejected on save. - Structure: stages (title, objective, talk track, watch-out, questions = the meeting checklist) · tones (name, when, delivery) · keyword triggers ("deixas") · attachments (existing http(s) URLs) · guardrails (what the copilot must never suggest). Limits: 20 stages, 30 questions/stage, 12 tones, 40 triggers, 12 terms/trigger, 30 attachments, 20 guardrails.
- Save is WHOLE-DOCUMENT (
playbooks_update= PUT). Omitted collections are wiped, attachments included, and they cannot be re-uploaded through this surface. Always read-modify-write fromplaybooks_get. - Ids (stage, question, tone, trigger) are authored by the caller (1-40 chars). Questions must be unique across the WHOLE playbook, not per stage: the room marks a question as covered by id. Keep existing ids stable on edits; a changed id is a different item.
- Keyword triggers are deterministic and model-free: the room compares each GUEST utterance with the terms (case, accents and punctuation ignored, whole words only:
carofires on "ficou caro demais", not on "carro alugado"), shows the same cue text instantly, and the same trigger returns at most every 5 minutes. Only the guest speaking fires them. Terms need at least 2 chars. - Editing a playbook changes meetings booked AFTER the save and future calls. A booked meeting keeps the script it received; the host refreshes it by re-picking the playbook inside the room.
- Delete is hard and permanent. Pipelines that suggested it are left with no suggestion; booked meetings keep their room script.
- Writes (create/update/delete/link/simulate) are for owner/admin/editor; every role can read.
- The talk track accepts 6000 chars but the meeting panel shows the first 2000.
playbooks_meeting_guides_getshows the truncated, real version. - Not available here: audio transcription for the simulator, file upload for attachments, AI-generated playbooks ("Com IA" in the editor), templates gallery (NEPQ, SPIN, BANT, Challenger, consultative diagnosis are editor templates, not API objects).
Thought process
- Find out what the user has:
playbooks_listfirst; edit before creating a near-duplicate. - New method from scratch: interview briefly (sales motion, buyer, stages the conversation goes through, top objections, forbidden promises), then design; do not invent pricing or promises the user never gave.
- Separate three jobs: authoring the document, linking it (suggestion per pipeline), choosing it for one meeting.
- Validate a draft before or right after saving with the rehearsal; treat its output as one sample, not a verdict.
Execute guide
- Build: draft
stages(one clear objective each, 3-8 open questions that steer the buyer, a short talk track, a concrete watch-out), 2-5tonesfor moments such as opening/objection/closing,keywordsfor predictable moments (price, competitor, "preciso pensar") andguardrailsfor hard "never" rules. Thenplaybooks_createwithscopes: ["meeting"]and[]for collections you have nothing for. Generate ids likestage-1,q-1-1,tone-1,cue-1. - Edit:
playbooks_get-> change only the intended items ->playbooks_updatewith the full document, ids preserved. - Make pipelines follow it:
playbooks_pipelines_listshows every pipeline withlinkedandcurrentPlaybookName. Attaching replaces the pipeline's current suggestion, so confirm whencurrentPlaybookNameis not null. Thenplaybooks_pipelines_attach/playbooks_pipelines_detach. From the pipeline side usepipelines_playbook_get(returns the playbook and the default copilot agent) andpipelines_playbook_set: absent field = untouched,null= clear;defaultCopilotAgentIdis a separate concern from the method and no tool here lists AI agents, so only set it with an id the user gives. - One meeting:
playbooks_meeting_guides_listgives the choices (refis the playbook id); pass it asplaybookIdonscheduled_calls_create, or change it withscheduled_calls_update(null= no playbook, omitted = keep, and creating without it falls back to the pipeline suggestion). - Preview:
playbooks_meeting_guides_getreturns what the room receives (panel guide, narrated script, guardrails). - Rehearse:
playbooks_simulateruns the real copilot on a made-up conversation. It has no playbook id: pass the method throughbriefing. For a saved playbook takescriptandguardrailsfromplaybooks_meeting_guides_get; addbriefing.goalandbriefing.persona. Send the wholeturnslist each call (host/guest); a guest line as the last turn mimics how the room wakes the copilot. Varyreason(objection,question,silence...) to probe stages. Use 3-6 scenarios at most (each is a real AI call). - Trigger check: the simulator does not run keyword triggers. To verify one, compare the sample sentence with the trigger terms by hand using the matching rule above and say so.
- Delete:
playbooks_list-> readpipelinesCount-> state how many pipelines lose their suggestion ->playbooks_deleteonly after explicit confirmation.
Report
- Lead with the job: created / edited / linked / previewed / rehearsed / deleted, then the playbook name.
- Created or edited: stages count, questions count, tones, triggers, guardrails, scopes; on edits list what changed and what was kept.
- Link changes: pipelines added/removed by name, plus any suggestion that was replaced.
- Rehearsal: per scenario show the last guest line, then the suggestion (
title,text,tone) or "no suggestion" withreason; separate "the copilot chose silence" from "the AI call failed" (reasonstarting withai_action_). End with what to tweak in the playbook, offered as opt-in. - Cap long lists with
+N more; follow-up mutations are opt-in.
Warnings
playbooks_updatewithout the full current collections silently deletes them; never send a partial document.- Attaching to a pipeline that already suggests another playbook replaces it without asking the pipeline's owner.
- Deleting a playbook cannot be undone.
- Rehearsal answers vary between calls and cost AI calls; one silent answer is normal (most turns need no card) and does not mean the playbook is broken.
- Do not promise that editing fixes meetings already booked.
Anti-patterns
- Treating the playbook as a pipeline setting, or trying to create one per lead.
- Regenerating ids on every edit.
- Writing prices, discounts or deadlines into stages or triggers that the user never provided.
- Marking
whatsapp/flowsscopes. - Looping the simulator to "optimize" the text without the user asking.
- Setting the pipeline's default copilot agent as a side effect of linking a playbook.
- Creating a playbook to configure an AI SDR agent, or saying a playbook was linked/associated to an agent: no such link exists.
defaultCopilotAgentIdonly picks the meeting copilot; say so, and point to the agent persona for what it should ask.
Clickmax skill revision: 23f2b5440092
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/docs.clickmax.io/clickmax-playbooks">View clickmax-playbooks on skillZs</a>