Role Launch

---
name: metaview-role-launch
description: >
  Launch a new role end-to-end in Metaview from a single intake conversation — the flagship
  workflow that chains Notetaker, note templates, Sourcing, Application Review, research
  mode, and Reports. Trigger on "we just opened a new role, set everything up", "launch the
  [role] search", "turn this intake call into a live pipeline", "new req kickoff", "build the
  interview kit and start sourcing", or after a role intake/kickoff call happens. Produces a
  role brief and interview plan, stage-specific note templates (optionally mirroring your ATS
  scorecards), a calibrated sourcing search, an inbound Application Review ICP, a target-company
  research thread, and a tracking report — then a weekly pipeline pulse to keep it honest.
---

# Metaview Role Launch

The gap between "the HM told me what they want" and "a pipeline is running against exactly that" is usually two weeks of fragmented setup. This skill closes it in one sitting: everything the intake call contains becomes live infrastructure — interview plan, note templates, sourcing search, inbound triage, tracking — all consistent with each other because they come from the same source of truth.

**Who it's for:** recruiters and TA teams opening a new req; agencies launching a new search.
**Metaview products needed:** AI Notetaker (intake + interviews). Each downstream phase needs its product: Sourcing, Application Review, Reports. **The skill degrades gracefully — build what the workspace has, skip what it doesn't, and say so.**

## Why this is powerful

The two weeks after an intake call are where searches quietly go wrong: the JD says one thing, the interviewers assess another, sourcing calibrates to a third, and nobody notices until the debrief. Here the hiring bar is captured once — from the HM's own words — and every artefact derives from it: templates, search ICP, inbound review, tracking. Setup that takes a coordinated week happens in one sitting, and consistency isn't a discipline problem anymore, it's structural.

## Operating rules

- Tool names may carry a connector prefix; match on the trailing name. No internal UUIDs in user-facing output.
- **One plan, one confirmation.** Present the full launch plan (Step 2) and get a single go-ahead for the package. Re-confirm individually only for live/irreversible actions: replacing an existing Application Review ICP, posting to the ATS, or emailing candidates.
- Everything generated must trace to the intake source. Where the intake was silent (comp band, interview stages), ask — never invent the hiring bar.
- Name everything you create consistently: "<Role title> — <artefact>" so the team can find the set.

## Workflow

### Step 1 — Capture the source of truth

`get_user_context` once. Then locate the intake:

- **Best**: a recorded intake/kickoff call — `search_conversations` (conversation type "Role Intake", or search by the role/HM in recent event titles), pull `default:transcript`.
- **Good**: a JD plus 10 minutes interviewing the user for what JDs never say: must-haves vs nice-to-haves ranked by emphasis, seniority band, comp, target companies, red flags, interview stages and owners, success metrics for the first year.
- If the intake call exists but is thin, read it AND ask the gap questions.

Distil the **role brief** — the single artefact every downstream phase derives from:

> Role summary · why this hire, why now · must-haves (testable, ranked) · nice-to-haves · red flags · target/feeder companies · comp band · location/visa · interview process (stages, owners, what each stage tests) · first-year success criteria

Show it to the user, take corrections. This is the moment to get it right.

### Step 2 — Propose the launch plan

Based on what the workspace has (check quietly: `list_note_template_groups`, `list_sourcing_searches`, `list_application_reviews(search_term=<role>)`, `list_fields` for reusable AI fields), propose the package, e.g.:

> 1. Interview plan + stage note templates (screen, HM, panel) — mirroring your Greenhouse scorecards
> 2. Sourcing search, ICP from the brief, first pack for calibration
> 3. Application Review ICP for inbound applicants
> 4. Research thread: target-company map for outreach
> 5. Tracking report + weekly pulse
> Say go, or tell me which to cut.

Flag anything that already exists (a search or review for this role) and propose refining it instead of duplicating.

### Step 3 — Interview kit and note templates

1. From the brief, draft the interview plan: per stage — what it assesses (mapped to must-haves), suggested questions (behavioural + situational, role-specific, no generics), and who owns it.
2. Create a note template per stage with `manage_note_template(action="create")`. Section design rules: each section's `description` is a complete instruction (scope + format contract + "only include what was actually discussed — do not fabricate; write 'Not discussed' otherwise"); sections for the stage's competencies plus Motivations and Housekeeping (comp, notice, location, right-to-work) on screens; `output` type TEXT with `text_format` BULLET_POINT for evidence sections. Non-evaluative phrasing throughout — sections capture what was said; the interviewer adds their own assessment.
3. **ATS scorecard mirror (high-leverage option):** if the team submits scorecards in the ATS, fetch a completed example from this role's interviews (`fetch_candidates` → scorecards) or ask the user to paste the scorecard form, then mirror its exact headings as template sections so interviewers can transfer notes field-by-field — objective sections filled from the transcript, subjective ones ("your recommendation") left to the human with a one-line evidence recap.
4. Place templates in the right public folder (`list_note_template_groups`, then `group_name` on update); tell the user to set when-to-use defaults in the Metaview app if they want automatic application per stage.
5. **Validate on a real call** if any comparable interview exists: `generate_notes` with `template="custom_definition"` + the new template on one conversation, read the output, tighten section descriptions once. Updating a template REPLACES all sections — always send the full section list on update.

### Step 4 — Sourcing search

Compose the kickoff brief from the role brief (mission, must-haves, strong signals, seniority, targets, exclusions, "a profile meeting these criteria looks like / a near-miss on these criteria looks like") and `send_sourcing_message(mode="source")`. It responds async (minutes): continue with Steps 5-7, then come back, present the proposed ICP to the user, and lock it. Full calibration discipline lives in **metaview-sourcing-copilot** — offer to run the first candidate-review session when the first pack lands.

### Step 5 — Inbound triage (Application Review)

If the workspace has Application Review and the role takes inbound: `list_application_reviews(search_term=<role>)`.

- Review exists → `get_application_review_details`, compare its active ICP to the role brief, and propose an aligned revision if they disagree. `update_application_review_icp` is live and reranks immediately — show full before/after, get explicit approval, pass `base_version`.
- None exists → reviews are created in the Metaview app; give the user the ICP text to paste (derived from the brief: role summary, must-haves, nice-to-haves, red flags) and pick up triage in **metaview-sourcing-copilot** once it's running.

### Step 6 — Target-company research (optional, runs in background)

`send_sourcing_message(mode="research")`: feeder-company map for this role/geo — who has the talent, who's a likely source right now (layoffs, post-acquisition), comp reality check. Takes 10-30 minutes; deliver as a follow-up and feed the company list back into the search ICP. Detail in **metaview-talent-research**.

### Step 7 — Tracking: report + weekly pulse

1. Reuse-first: `list_fields(search_term=...)` for existing fields (comp expectations, reasons for leaving, competitor mentions) before creating role-specific ones. Only add fields this role genuinely needs (e.g. a must-have coverage check on screens); confirm before creating; calibrate per **metaview-interview-intelligence** before trusting.
2. `create_report` (confirm first): filters = this role's conversations (job/application field where ATS-synced, else event-title pattern + date), columns = date, candidate, interviewer, stage, scorecard recommendation + the AI fields. Share the URL.
3. **Weekly pulse** (offer as a scheduled task, or run on demand): interviews happened + scheduled, pipeline movement by stage, sourcing calibration state (`get_search_details`: acceptance rate, unreviewed count), new closely-matching applicants, what candidates are saying (comp vs band, recurring concerns, decline reasons), and the one thing to fix this week. Post to chat or the team's channel.

### Step 8 — Hand over the launch pack

Close with a compact summary: the role brief (final version), what was created (each artefact by name, with URLs where they exist), what's still processing (search ICP, research thread) and when to check back, and the owner actions (set template defaults in-app, create the Application Review, book the first calibration session).

## Edge cases

- **No intake call and no JD**: ask the user the intake questions yourself (the gap questions in Step 1) — 10 focused minutes beats launching on guesses. Offer the role's built-in `intake_call` template for the *next* HM conversation so it's captured properly.
- **Relaunching a stalled role**: same skill, but start by reading what exists (old search's feedback history, the report's AI-field signal, past-stage interviews) and lead the plan with "what changed".
- **Agency context**: the "intake" is the client kickoff call; add a client-facing variant of the brief and weekly pulse (internal notes stripped).
- **Workspace missing a product** (no Sourcing, no AR): build the rest; name what was skipped and why, without selling.

## Example invocations

- "We just did the intake for a Senior Platform Engineer in Warsaw — set the whole thing up."
- "Launch the Engagement Manager role: templates mirroring our Ashby scorecard, search, and a tracking report."
- "Relaunch the Head of Finance search — it's been stuck for a month."
- "Here's the JD for our first US AE. Build the kit, start sourcing, and give me a Monday pulse."