跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
营销与销售

pm-gtm-v1

Go-to-market and growth planning, GTM here means go-to-market, never a tag-manager container. Use when a growth plan needs building or refreshing: "build the go-to-market plan", "growth plan for <period or launch>", "campaign architecture", "channel plan", "media plan", "tracking planning rows", "update the growth plan after <release>". It resolves which plan document is actually current, takes a FRESH gate-state snapshot before planning a single sentence, builds the plan phase by phase, narrates the campaign architecture per channel and per persona, and emits channel and tracking rows CONFIG-ONLY, no credentials, no platform writes, no spend. NOT for executing campaigns or buying media; NOT for implementing tracking (that is engineering work, routed through pm-requirements-v1); NOT for writing a handoff package or scoring a backlog.

naderelewa · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add naderelewa/Product-to-Prod --skill pm-gtm-v1

技能说明


name: pm-gtm-v1 description: | Go-to-market and growth planning, GTM here means go-to-market, never a tag-manager container. Use when a growth plan needs building or refreshing: "build the go-to-market plan", "growth plan for ", "campaign architecture", "channel plan", "media plan", "tracking planning rows", "update the growth plan after ". It resolves which plan document is actually current, takes a FRESH gate-state snapshot before planning a single sentence, builds the plan phase by phase, narrates the campaign architecture per channel and per persona, and emits channel and tracking rows CONFIG-ONLY, no credentials, no platform writes, no spend. NOT for executing campaigns or buying media; NOT for implementing tracking (that is engineering work, routed through pm-requirements-v1); NOT for writing a handoff package or scoring a backlog. argument-hint: "<gtm goal / 'refresh the growth plan'> [--channels <list|propose>] [--budget-posture <direction-only|capped|none>] [--stack ]" user-invocable: true

pm-gtm-v1, source of truth → fresh gate state → phase-gated plan → config-only rows

Input is a go-to-market goal: a launch, a period, a plan refresh. Output is a plan package: a fresh gate-state snapshot, a plan built phase by phase, a campaign-architecture narrative, channel and tracking planning rows, and a handoff record for the owner and whoever runs marketing operations.

This verb plans. It never spends, never posts, never configures a platform, and never touches a tag container. Execution stays human; implementation stays engineering's.

Seam discipline: this verb emits no engineering seam artifacts. When technical work emerges from planning, a pixel install, a container change, instrumentation, landing pages, it leaves as classified candidate rows routed to /pm-requirements-v1 (a single item) or /pm-portfolio-v1 (a batch). One seam, one owner.

Preflight (run first)

PKG_ROOT="${CLAUDE_PLUGIN_ROOT:-${PKG_ROOT:-}}"
[ -n "$PKG_ROOT" ] && [ -r "$PKG_ROOT/.claude-plugin/plugin.json" ] || { echo "STOP: set PKG_ROOT to this package's root directory, the one holding .claude-plugin/plugin.json, then re-run."; exit 2; }
bash "$PKG_ROOT/scripts/preflight.sh" pm-gtm-v1 [--stack <keys>]

Conditional keys: gate-state-snapshot (the read-only analytics reads that let a gate state be measured rather than asserted) · workspace-docs · backlog-sync. A selected-conditional miss blocks like a required one.

Standalone vs supercharged. Standalone (python3): the full chain resolve, the phase-gated plan, the campaign narrative and the config-only rows, with every unprobeable gate state written NEEDS-CONFIRMATION with a named confirmer, never guessed. Supercharged: gate-state-snapshot turns those rows into measured states.

There is deliberately no advertising-platform capability. Nothing here holds or requests ad-account credentials, which is precisely why the rows in step 4 are configuration proposals rather than changes.

Hard rules

  1. Resolve the source-of-truth chain first, and never plan from stale content. Plan documents accumulate: an update note usually rides on top of an older attested canonical. Start from the newest update document; use the canonical for constants and structure only. Every gate state embedded in an older plan is frozen to the date that plan was written, treating those as current is the specific failure this verb exists to prevent.
  2. A fresh gate-state snapshot precedes the first planning sentence. The snapshot is written and presented at gate 1, it is the ONLY gate-state source for the whole run, and it expires when the run closes. The next run re-snapshots; it never inherits.
  3. Budget language is posture-bound. direction-only is the default and never invents a per-channel cap. capped uses only a number the owner named, entered as an [L0] constant. none means planning without spend. Never convert direction into numbers, and never treat a stated cap as a pool to allocate, per-spend approval survives every posture.
  4. The rows are config-only. No credential file is read (pointers and names only), no platform API is called, no pixel, tag, container or campaign is created or edited, and nothing is verified against a live ad account. Every row carries state: CONFIG-PLANNED and, where it applies, what blocks it.
  5. No seam artifacts from this verb. Emerging technical work becomes candidate rows pre-classified risk_class / money_path / analytics_touch, routed to the verb that owns the seam. Go-to-market is not tag-manager work: container changes are engineering, and the shared acronym must never pull them in here.
  6. Customer-facing claim safety. Build the unsupportable-claims list at lock, one row per claim the business cannot substantiate, with a named reviewer per locale, and bind it to every headline and hero surface from the first draft, not just at the final audit. Planning rows and creative-brief pointers are customer-facing-adjacent surfaces.
  7. Honesty on gates. A red gate stays red until its owner reports it unblocked. A gate that cannot be probed is NEEDS-CONFIRMATION with a named confirmer, never a guessed state, and never quietly softened at handoff.
  8. No volume targets. No channel quotas, no row counts as goals. Deliverables are versioned forward, never overwritten; prior plan epochs are read-only [L4] archive.
  9. Artifacts land where the run said they land. Write every emitted file under the folder this run selected; with none selected, the session's already-authorized working path; with no authorized writable path at all, STOP and ask for one before any artifact is written. The harness enforces that boundary either way, the ask is what turns a refusal into a decision.

Inputs

| Input | Required

阅读完整来源 ↗