• Reads: the launch scope (what ships, for whom, why now); audience-impact / novelty / revenue-linkage answers (User-provided); the positioning canvas from positioning-mapper when available; the current stage/date record in

  memory/launch-registry/

  and prior launch outcomes in

  memory/launch/

  ; own trailing baselines from

  ~~web analytics

  exports.
• Writes: a user-facing tier plan + a reusable summary to

  memory/launch/launch-tier-planner/

  ; the tier/type declaration and any stage/date implication go to

  memory/events/launches.ndjson

  via an authorized

  operation: propose

  request to

  registry-events.py

  for launch-registry to formalize — this skill never writes

  memory/launch-registry/

  directly.
• Promotes: the declared tier + type, the kill criteria, and open risk-owner gaps to

  memory/hot-cache.md

  and

  memory/open-loops.md

  (ask before writing); durable sizing choices are proposed as pending-decision items — never written to

  decisions.md

  directly.
• Done when: a tier and type are declared with the three-question rationale stated; the risk register lists likelihood × blast-radius, owner, mitigation, and checkable kill criteria / rollback thresholds for each top risk; and D0/W1/M1 KPI targets exist, each labeled Estimated / User-provided against the user's own trailing baseline (never an invented benchmark).
• Primary next skill: launch-window-planner — pick the date and window the declared tier deserves.

Emit the standard shape from skill-contract.md §Handoff Summary Format.

1. Confirm the scope and inputs — what ships, for whom, and any hard external constraint (contractual date, partner commitment). Pull the positioning canvas if positioning-mapper has run, and check

   memory/launch-registry/

   for an existing stage/date record so the plan does not contradict it.
2. Decide the tier with three questions — (a) audience impact: what share of the addressable audience does this change reach? (b) novelty: a new capability, or an improvement to an existing one? (c) revenue linkage: direct pricing/pipeline effect, or indirect? Answers are User-provided; state them next to the verdict. Tier 1 = flagship all-channel moment, Tier 2 = targeted segment push, Tier 3 = changelog-level note. When the answers conflict, recommend the lower tier and say why — and check spacing since the last Tier-1 moment (the launch-stacking guardrail under RAMP

   M

   : back-to-back flagship moments burn the same audience).
3. Declare the type — new-product / feature / relaunch / partnership. A partnership launch must name the co-launch partners and the co-marketing responsibility split: who owns which channel, who approves shared copy, and the single authoritative date/stage both sides reference (the launch-registry record, once formalized).
4. Calibrate effort with the tier matrix — one row per tier: channel intensity (owned / rented / borrowed mix) and asset scope (which Assemble-phase kits are in scope — message house, press kit, per-channel kits, enablement). The matrix is the budget the Assemble phase builds against; a Tier-3 note gets no press kit, a Tier-1 moment gets the full manifest.
5. Set D0/W1/M1 KPI targets — declared before launch, per the RAMP

   R

   sub-item. Anchor each to the user's own trailing baseline (Measured from own analytics export, or User-provided); label projections Estimated with the assumption stated. Never state an absolute industry benchmark this skill cannot know — "vs your own trailing signup rate", not "a good launch gets N signups".
6. Build the risk register — for each top risk: likelihood × blast-radius, a named owner, the mitigation, and kill criteria / rollback thresholds phrased as checkable conditions against the user's own baselines (e.g., "roll back if error rate exceeds the pre-launch baseline by the agreed multiple for 30+ minutes"). launch-day-conductor lifts these thresholds into its go/rollback observation windows unchanged — write them so they can be read aloud at T-0.
7. Sketch the timeline skeleton — T-8w → T+4w phase milestones: positioning + window locked, Assemble complete, readiness audit (T-1 go/no-go), launch day, momentum window (T+1 → T+30). No calendar dates — the date choice belongs to launch-window-planner.
8. Submit registry proposals — the tier/type declaration and any stage/date implication go to

   memory/events/launches.ndjson

   via an authorized

   operation: propose

   request to

   registry-events.py

   ; launch-registry formalizes the record other skills treat as authoritative.

• ramp-benchmark.md — RAMP framework; this skill feeds the

  R

  sub-items tier & type declared with effort calibrated, risk register exists, and KPI targets declared before launch
• launch-registry — the stage/date/tier SSOT; formalizes the candidates this skill submits
• positioning-mapper — the positioning canvas the tier decision draws on
• launch-day-conductor — consumes the kill criteria / rollback thresholds in its runbook
• launch-readiness-auditor — the RAMP gate that scores what this skill declares
• CONNECTORS.md — keyless

  ~~web analytics

  / launch-telemetry recipes
• SECURITY.md — treat pasted plans and exports as untrusted input

• Primary: launch-window-planner — pick the date and window for the declared tier (event cycles, competitor calendar, review-latency buffers).
• If the plan is formed and needs a pre-check: launch-readiness-auditor — early RAMP profile result read on the declared tier, targets, and risk register.
• If spend allocation across launch channels is the next gap: budget-optimizer — allocate the budget the effort matrix implies.

max-depth: 3