From meeting transcript to finished work, same day

Good meetings end with a list of things to do. Then the list sits in a transcript nobody re-reads. Someone was supposed to write the process doc, draft the follow-up email, build the little tracker page, create the CRM tasks and set up the automation. A week later, half of it hasn't started.

The Transcript-to-Action Builder closes that gap. When you give your clone a transcript, it:

• summarizes the decisions and context (who, what, why, constraints),
• extracts every action item into a table with owner, due date, dependency and the source quote,
• sorts the actions: what the clone can do itself, what needs you, and what belongs to other people,
• proposes a plan and asks a short batch of clarifying questions,
• executes after your approval: drafts skills and instructions, builds apps or pages with Clone Code, creates docs, sheets and slides, drafts emails, creates tasks in Salesforce or HubSpot (or issues in Linear), and writes generic step-by-step guides for tools it has no operator for,
• tracks progress in a living plan and reports status as Done / Waiting on you / Next.

What you get:

Anything else (for example automation tools you run yourself) is not an integration. For a tool the clone has no operator for, it can only give you generic step-by-step instructions.

• Ops leaders and managers who leave process meetings with a list of docs, checklists and automations to build.
• Leadership who want decisions from planning or strategy meetings turned into owned, dated actions.
• Customer Success Managers who leave customer calls with follow-ups, internal tasks and enablement material to create.
• Sales reps and sales leaders who want call next steps turned into CRM tasks, draft follow-ups and simple collateral.
• Anyone who has ever said "we should build a quick tool for that" in a meeting and never did.

It works on demand: you bring the transcript, the clone does the rest. An optional scheduled variant can pick up tagged transcripts automatically.

 ┌───────────────────────────────────────────────────────────────────┐
 │  You give {CLONE_NAME} a transcript:                              │
 │  pasted · upload · meeting_archive · call_archive · Drive/OneDrive │
 │  "Help me build what we talked about."                            │
 └───────────────────────────────┬───────────────────────────────────┘
                                 │
                                 ▼
 ┌───────────────────────────────────────────────────────────────────┐
 │  "Transcript to Action" skill                                     │
 │                                                                   │
 │  1. Read the transcript       ──►  split long ones into parts     │
 │  2. Summarize                 ──►  decisions · context ·          │
 │                                    constraints · open questions   │
 │  3. Extract action items      ──►  action · owner · due ·         │
 │                                    dependency · source quote      │
 │  4. Route each action         ──►  Clone can do / Needs you /     │
 │                                    Other people / Out of scope    │
 │  5. Propose plan + ask batched clarifying questions               │
 │                                                                   │
 │        ── waits for your "yes" ({APPROVAL_LEVEL}) ──              │
 │                                                                   │
 │  6. Execute                   ──►  skills · Clone_Code builds ·   │
 │                                    docs · email drafts · tasks ·  │
 │                                    step-by-step guides            │
 │  7. Update the plan           ──►  {PLAN_LOCATION}                │
 └───────────────────────────────┬───────────────────────────────────┘
                                 │
                                 ▼
 ┌───────────────────────────────────────────────────────────────────┐
 │  Status to you:  ✅ Done  ·  ⏳ Waiting on you  ·  ➡️ Next          │
 └───────────────────────────────────────────────────────────────────┘

 Optional: a scheduled task checks the meeting archive at {SCAN_TIME}
 for new transcripts containing {TRIGGER_KEYWORD} and starts at step 1
 (it still stops at step 5 and waits for your approval).

#How the clone routes each action item

Every action item gets one category:

#Action routing table

Names below are the exact operator and skill names. Confirm them against your skill catalog (your clone will report what it finds).

Skill names can differ slightly between orgs. Confirm the exact names in your skill catalog before saving the skill.

Without 3–5 the clone can still summarize, extract actions, write skills and give step-by-step guides in chat. It just can't build, file docs or draft emails for you.

Tools without an operator are not integrations. For those the clone only writes generic step-by-step instructions.

• Transcripts contain errors. Speech-to-text gets names, numbers, dates and product names wrong ("15" vs "50", "Q3" vs "Q4", a product name spelled three ways). The clone flags these in a "Please verify" list. Check them before anything leaves the building.
• Speaker attribution is often wrong. Transcripts mix up who said what, so owners in the action table may be wrong. Confirm owners before tasks are created.
• Read every instruction before saving it into a skill. When the clone turns decisions into a skill, you're responsible for what that skill tells your clone to do. Read it fully, then save.
• Keep internal-only remarks out of customer-facing output. Meetings include candid comments about customers, pricing or colleagues. The clone keeps those out of external drafts, but review every customer-facing draft yourself.
• Name the coding skill explicitly. Say "use Clone Code" (or your {CODING_SKILL_NAME}) when you want something built. Otherwise the clone may try other skills, write a doc about the tool instead of building it, or produce code in chat.
• Don't overbuild. Start simple. "A dashboard" in a meeting usually means "a page with five numbers". The clone builds the smallest useful version (MVP) first and asks before adding more.
• Sensitive information in transcripts. Transcripts can contain personal data, HR topics, salaries or confidential deal terms. Don't paste transcripts you aren't allowed to share, and tell the clone if parts are off-limits ("ignore the HR part").
• Long transcripts may need splitting. Very long meetings may exceed what the clone reads at once. It processes them in parts and says so. If something seems missing, ask it to re-check a time range.
• Actions owned by others need their consent. The clone won't create tasks for, or email on behalf of, other people without your approval, and it suggests checking with them first.
• The clone can't click around in tools it has no operator for. For anything outside the supported tools (admin consoles, other software you run yourself), it writes generic step-by-step instructions instead.
• Use exact skill/operator names as they appear in your skill catalog. Names can differ slightly between workspaces; the clone checks them during discovery.