[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M75DXB07YPGWNENXW5MM9RR7":3,"skill-same:01M75DXB07YPGWNENXW5MM9RR7":59},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":37,"commitSha":38,"skillDirectory":39,"forksCount":40,"repositoryHtmlUrl":41,"repositoryHomepageUrl":42,"ownerHtmlUrl":43,"ownerBlogUrl":44,"ownerDescription":6,"updatedAt":45,"install":46,"resources":58},"01M75DXB07YPGWNENXW5MM9RR7","engagement-workflow",null,"Run a full marketing engagement using the 12-Part methodology. Use when starting a new engagement, advancing parts, applying the Decision Matrix, or showing engagement status.","LongLeo287\u002FSEOSONA-OS","LongLeo287","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F141803367?v=4","01M6JK75C62VCNZKR5BRZ386YV","LongLeo","01M75CK60XQGYJY3YDGGFJQDYT",2,"MIT","https:\u002F\u002Fgithub.com\u002FLongLeo287\u002FSEOSONA-OS\u002Fblob\u002Fmain\u002F2_KNOWLEDGE\u002Fframeworks\u002Fdigital-marketing-pro\u002Fskills\u002Fengagement-workflow\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: engagement-workflow\ndescription: \"Run a full marketing engagement using the 12-Part methodology. Use when starting a new engagement, advancing parts, applying the Decision Matrix, or showing engagement status.\"\nuser-invocable: true\ntriggers:\n  - start a new engagement\n  - run the 12-part methodology\n  - advance engagement to next part\n  - show engagement status\n  - apply the decision matrix\n  - re-run v2 documents\n  - mark engagement part complete\n  - what part of the engagement are we on\nallowed-tools: Read Write Edit Bash Glob Grep\nengagement-part: orchestrator\nview-preference: both\n---\n\n# \u002Fdigital-marketing-pro:engagement-workflow — 12-Part Engagement Orchestrator\n\nThis skill orchestrates the full marketing engagement using the 12-Part sequential methodology. Every brand engagement runs through the same 12 parts in sequence, producing a canonical set of files at each stage.\n\n## Context efficiency\n\nHeavy skill. **Grep before Read** any referenced file, then `Read` only matched ranges with `offset` + `limit`. List the brand's workspace at `~\u002F.claude-marketing\u002Fbrands\u002F{slug}\u002F` (or `$CLAUDE_PLUGIN_DATA\u002Fdigital-marketing-pro\u002Fbrands\u002F{slug}\u002F` when that env var is set) before opening files. On re-invocation mid-session, skip files already in context.\n\nRead these references before producing output:\n- [engagement-flow-methodology.md](..\u002Fcontext-engine\u002Fengagement-flow-methodology.md) — the full 12-Part flow\n- [two-views-model.md](..\u002Fcontext-engine\u002Ftwo-views-model.md) — v1 \u002F v2 architecture\n- [stone-vs-opinion.md](..\u002Fcontext-engine\u002Fstone-vs-opinion.md) — confidence tagging\n- [decision-matrix-rerun.md](..\u002Fcontext-engine\u002Fdecision-matrix-rerun.md) — when to re-run what\n- [update-back-rule.md](..\u002Fcontext-engine\u002Fupdate-back-rule.md) — versioning protocol\n- [living-instruction-file-spec.md](..\u002Fcontext-engine\u002Fliving-instruction-file-spec.md) — LIF schema\n\n## Operating Mode\n\nThis skill is invoked via the `\u002Fdigital-marketing-pro:engagement` command family. The command is a thin router — **this skill is the single source of truth** for the engagement lifecycle, the checkpoint protocol, and the per-part production contract. Each subcommand maps to a specific lifecycle action. The skill calls `engagement-state.py` for persistence via:\n\n```\npython \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fengagement-state.py\" \u003Csubcommand> ...\n```\n\nYou should never hand-edit `_engagement.json` — always go through `engagement-state.py`.\n\n## Checkpointing & Resume (single source of truth)\n\nEvery long engagement run is resumable. The checkpoint protocol is: **init a run → save each part as it completes → finalize → publish to the visible output folder.** This lets an interrupted run (context exhaustion, user cancel, machine sleep) resume from the next un-checkpointed part instead of restarting from Part 1.\n\n**1. On `start`, after the brand pre-condition passes, open a checkpoint run and link it to engagement state:**\n\n```bash\npython \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fcheckpoint-manager.py\" init \\\n    --brand \"{brand_slug}\" --workflow engagement --topic \"{engagement_id}\"\n\n# Record the returned run_id into _engagement.json so resume can find it:\npython \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fengagement-state.py\" set-checkpoint-run \\\n    --brand \"{brand_slug}\" --id \"{engagement_id}\" --run-id \"{run_id}\"\n```\n\n`set-checkpoint-run` stores the run_id in `_engagement.json`, making the resume linkage real (previously the run_id was never persisted).\n\n**2. After each part completes and passes its quality gate, the orchestrator saves that part's output:**\n\n```bash\npython \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fcheckpoint-manager.py\" save \\\n    --brand \"{brand}\" --run-id \"{run_id}\" \\\n    --step {part_number} --content-file \"{path_to_that_part_deliverable}\" --extension md\n```\n\nPass the **actual deliverable path for that part** (e.g. Part 3 saves the Four Core Documents path; Part 8 saves the Growth Plan path) — never a placeholder for a different part.\n\n**3. Before saving Part 5 (Client Validation) and Part 8 (Growth Plan) deliverables, run the full quality gate:**\n\n```bash\n# BLOCKING gate — Part 5 and Part 8 deliverables cannot be checkpointed until this passes\n\u002Fdigital-marketing-pro:check \"{path_to_deliverable}\" --full --brand {brand}\n```\n\nIf `\u002Fdigital-marketing-pro:check --full` returns BLOCKED, fix the CRITICAL issues before checkpointing the part.\n\n**4. After the final part, publish every artifact to the user-visible folder and finalize:**\n\n```bash\npython \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Foutput-publisher.py\" publish-run \\\n    --brand \"{brand}\" --run-id \"{run_id}\"\n\npython \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fcheckpoint-manager.py\" finalize \\\n    --brand \"{brand}\" --run-id \"{run_id}\" --status completed\n```\n\nThen point the user at the visible output folder via `\u002Fdigital-marketing-pro:output-folder {brand}`.\n\nTo resume an interrupted run, use `\u002Fdigital-marketing-pro:resume` — it reloads every saved part and continues from the next un-checkpointed part.\n\n## State validation & rework caps\n\n- **Validate a part's outputs against the manifest** before marking it complete:\n  ```bash\n  python \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fengagement-state.py\" validate-part \\\n      --brand \"{brand}\" --id \"{id}\" --part {N}\n  ```\n  This diffs the actual files on disk against the `PART_DEFINITIONS` manifest and flags missing deliverables. Use it in `file-tree` and before `next`.\n\n- **Repair a partially-initialised engagement directory** (instead of crashing on a non-empty dir):\n  ```bash\n  python \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fengagement-state.py\" init --repair \\\n      --brand \"{brand}\" --id \"{id}\"\n  ```\n  `--repair` completes the canonical directory tree and state file on a dir that holds only partial state.\n\n- **v2 re-run cap:** a maximum of **2 v2 re-run rounds per part** is allowed without explicit user override. The round count is stored in `_engagement.json`. If a part would exceed 2 rounds, stop and ask the user to explicitly approve further re-runs (records the override in state). This prevents unbounded re-run loops.\n\n## Subcommands\n\n### `\u002Fdigital-marketing-pro:engagement start \u003Cbrand-slug> \u003Cengagement-id>`\n\n**Purpose:** Initialise a new engagement.\n\n**Steps:**\n\n1. Validate that the brand profile exists at `~\u002F.claude-marketing\u002Fbrands\u002F{brand-slug}\u002Fprofile.json`. If not, instruct the user to run `\u002Fdigital-marketing-pro:brand-setup` first.\n2. Run `python ${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fengagement-state.py init --brand {brand-slug} --id {engagement-id}`.\n3. Confirm the directory tree was created and report the next required action (Part 1 intake).\n4. Walk the user through Part 1 Stone vs Opinion intake by asking the questions one batch at a time.\n\n**Part 1 intake questions (ask in this order):**\n\n**Stone — what the client knows for certain:**\n\n1. Company basics: founded year, employee count, headquarters location, geographic operations\n2. Business model: revenue streams, pricing tiers, primary product\u002Fservice categories\n3. Current marketing: channels currently active, monthly marketing spend, current measurable KPIs\n4. Tech stack: CRM, email service provider, analytics setup, ad accounts\n5. Customer base scale: customer count, biggest named customer, average order value if known\n\nFor each Stone fact, capture:\n- The fact itself\n- Source (how the user knows \u002F what document confirmed it)\n\nSave each via:\n```\npython ${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fengagement-state.py add-stone-fact --brand {slug} --id {id} --fact-json '{\"category\":\"...\",\"fact\":\"...\",\"source\":\"...\"}'\n```\n\n**Opinion — what the client believes:**\n\n1. Brand positioning: how does the client describe their position in the market?\n2. Customer base: who do they think their customers are? Why do they buy?\n3. Competitors: who do they consider their main competitors?\n4. Growth opportunities: where do they think the biggest opportunity is?\n5. What is working: what marketing activity does the client believe is working?\n6. What is not working: what does the client believe is not working?\n\nFor each Opinion, capture:\n- The hypothesis\n- Client's evidence for it (could be intuition, anecdote, partial data)\n- Research question — what would the unbiased research need to verify or refute?\n\nSave each via:\n```\npython ${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fengagement-state.py add-opinion --brand {slug} --id {id} --hypothesis-json '{\"category\":\"...\",\"hypothesis\":\"...\",\"client_evidence\":\"...\",\"research_question\":\"...\"}'\n```\n\n**On completion of Part 1:** mark Part 1 as completed via `mark-part-completed --part 1`, advise the user to proceed to Part 2 (External Research).\n\n### `\u002Fdigital-marketing-pro:engagement next [brand] [id]`\n\n**Purpose:** Advance to the next part.\n\n**Steps:**\n\n1. Read engagement status via `engagement-state.py status`\n2. Identify the current part and next not-yet-completed part\n3. Confirm with the user that the current part is genuinely complete (do not auto-advance — ask)\n4. On confirmation, mark current as completed, advance current_part pointer\n5. Brief the user on what the new part requires\n\n### `\u002Fdigital-marketing-pro:engagement status [brand] [id]`\n\n**Purpose:** Show engagement status.\n\n**Steps:**\n\n1. Run `engagement-state.py status` — get the full state\n2. Read the Living Project Instruction File header\n3. Format a human-readable summary:\n   - Engagement: brand + id + start date\n   - Current part: part name + days in\n   - Completed parts: list\n   - Pending parts: list\n   - Open re-run decisions: count\n   - LIF last updated: date\n4. If the engagement has open items needing resolution, list them\n\n### `\u002Fdigital-marketing-pro:engagement file-tree [brand] [id]`\n\n**Purpose:** Show the engagement directory file tree.\n\n**Steps:**\n\n1. Run `engagement-state.py file-tree`\n2. Format as an indented tree\n3. Highlight files that are missing per the canonical structure. Use `engagement-state.py validate-part --part {N}` to diff each completed part's actual files against the `PART_DEFINITIONS` manifest (e.g., if Part 3 is marked completed but `3.1-business-and-sbu-analysis.md` is missing, `validate-part` flags it deterministically instead of eyeballing).\n\n### `\u002Fdigital-marketing-pro:engagement validate [brand] [id]`\n\n**Purpose:** Run the Part 5 Client Validation flow.\n\n**Pre-condition:** Parts 2, 3, 4 must be completed.\n\n**Steps:**\n\n1. Verify pre-conditions (Parts 2, 3, 4 completed)\n2. Invoke the `client-validation-document` skill — it produces the Part 5 deliverable: a structured document presenting each finding from v1 with ACCEPT\u002FREJECT\u002FEDIT\u002FDEFER options\n3. **Run the full quality gate on the Part 5 deliverable before it goes to the client:** `\u002Fdigital-marketing-pro:check \"{part5_path}\" --full --brand {brand}`. If it returns BLOCKED, fix the CRITICAL issues first (this gate is mandatory before Part 5 and Part 8 deliverables).\n4. After the user reviews and provides decisions, parse them into a triggers list per the Decision Matrix categories\n5. Run `engagement-state.py decision-matrix --triggers \"{comma-separated}\"` to compute the v2 re-run plan\n6. Present the re-run plan to the user\n7. Mark Part 5 completed; on user approval of the re-run plan, advance to Part 6\n\n### `\u002Fdigital-marketing-pro:engagement re-run-decision [brand] [id]`\n\n**Purpose:** Apply the Decision Matrix to compute v2 re-runs.\n\n**Steps:**\n\n1. Read the Part 5 Client Validation Document\n2. Categorise rejected\u002Fedited findings into Decision Matrix triggers\n3. Show the triggers and the computed re-runs\n4. Estimate the cost (rough token count) of each re-run\n5. Await user approval — they can accept, modify (skip some, add others), or reject\n6. Record the executed plan via `engagement-state.py record-rerun-execution`\n\n### `\u002Fdigital-marketing-pro:engagement update-back [brand] [id] --doc \u003Cdoc-id> --reason \u003Creason>`\n\n**Purpose:** Apply the Update-Back Rule to bump a source document version after Part 7+.\n\n**Pre-condition:** The user has already drafted the corrected document content.\n\n**Steps:**\n\n1. Read the current version of the doc\n2. Confirm the correction with the user (validation step per the Update-Back Rule)\n3. Bump the version via `engagement-state.py bump-version --doc {id} --reason \"{reason}\"`\n4. Save the new version file with a header noting v(prev) → v(new) changes\n5. Update the Living Project Instruction File via `lif-log-change` — it now appends the change to `living-instruction-file.md` and refreshes the header date, so the LIF reflects the correction immediately\n6. Identify downstream documents that may need review and add to the engagement's review queue\n\n### `\u002Fdigital-marketing-pro:engagement lif-show [brand] [id]`\n\n**Purpose:** Display the Living Project Instruction File.\n\n**Steps:** Run `engagement-state.py lif-show` and format the markdown output for readability.\n\n### `\u002Fdigital-marketing-pro:engagement list-engagements [brand]`\n\n**Purpose:** List all engagements (optionally filtered by brand).\n\n**Steps:** Run `engagement-state.py list-engagements --brand {slug}` and format as a table.\n\n### Production shorthands\n\nThe command family also exposes four production shorthands that route straight to the part-producing skills (documented in *Per-Part Production Targets* below). These match the command surface one-to-one:\n\n- `\u002Fdigital-marketing-pro:engagement four-core \u003Cbrand> \u003Cid> [--doc 3.X] [--view v2] [--combined]` — Part 3, invokes the `four-core-documents` skill\n- `\u002Fdigital-marketing-pro:engagement growth-plan \u003Cbrand> \u003Cid>` — Part 8, invokes the `growth-plan` skill\n- `\u002Fdigital-marketing-pro:engagement yearly-planner \u003Cbrand> \u003Cid>` — Part 8 companion, invokes the `yearly-planner` skill\n- `\u002Fdigital-marketing-pro:engagement loop \u003Cbrand> \u003Cid>` — Part 12, invokes the `continuous-improvement-loop` skill\n\n## Per-Part Production Targets\n\nEach part is produced by real, existing agents and skills. This orchestrator dispatches to the targets below — there are **no** wrapper skills named `external-research` \u002F `preparation-documents` \u002F `channel-strategy-fanout` \u002F `execution-artefacts` \u002F `ai-creative-instructions`; those never existed. Use the exact targets named here:\n\n| Part | Real target(s) |\n|------|----------------|\n| 1 | (this skill — intake walked here directly) |\n| 2 | agents `market-intelligence` + `competitive-intel`; skill `audience-intelligence` (invoke as a skill); reference `compliance-rules.md` (load as context) |\n| 3 | skill `four-core-documents` (produces 3.1, 3.2, 3.3, 3.4) |\n| 4 | command `competitor-analysis` + skills `audience-intelligence` + agent `market-intelligence` |\n| 5 | skill `client-validation-document` |\n| 6 | re-runs invoke skill `four-core-documents` with `--view v2` |\n| 7 | skills `content-engine` + `campaign-orchestrator` + `analytics-insights` |\n| 8 | skills `growth-plan` + `yearly-planner` |\n| 9 | per-channel skills — `paid-advertising`, `aeo-geo`, `social-strategy`, `seo-plan`, `email-sequence` (one per channel family) |\n| 10 | skill `content-engine` (execution \u002F output mode) |\n| 11 | skills `content-engine` + `ad-creative` + `video-script` (creative briefs); asset rendering happens in your own creative tooling (design team, AI image\u002Fvideo tools, or a connected design platform), then finished assets are signed via `c2pa-metadata` |\n| 12 | skill `continuous-improvement-loop` |\n\n## Parallel Dispatch\n\nSeveral parts of the engagement contain **independent sub-tasks** that should be dispatched **in parallel via multiple `Task` tool calls in a single message** — not sequentially. Dispatching independent sub-tasks concurrently is substantially faster than running them one after another; actual time varies by engagement depth, model, and rate limits. Keep concurrent subagents to a handful (roughly 3–8) — past that you queue against API rate limits and the win drops; under 3 there is nothing to parallelize.\n\n**Cost note:** total token usage is broadly similar (you're doing the same work) but billed-per-turn input costs trend up slightly because each parallel subagent re-loads its context.\n\n**Parts that benefit from parallel dispatch:**\n\n| Part | Parallel-eligible work | How to dispatch |\n|---|---|---|\n| **Part 2 — External Research** | Market sizing, competitor landscape, customer signals, regulatory landscape — none depend on each other | Dispatch the `market-intelligence` agent and the `competitive-intel` agent as parallel `Task` calls; invoke `audience-intelligence` as a skill; load `compliance-rules.md` (a reference file) as context — not as a subagent |\n| **Part 4 — Competitive + Customer + Market** | Four documents (4.1, 4.2, 4.3, 4.4) are independent — they reference Part 2 only | Dispatch all four in a single message with the four respective subagents |\n| **Part 9 — Channel Strategy Fan-out** | Up to 17 channel docs in 7 families. Families 2 (Paid platforms), 3 (Organic & Influencer), 4 (Marketplace & CRM), 5 (Content\u002FATL\u002FBTL\u002FPR) are independent after Families 1 (Search & Campaign) and 6 (Web + Measurement) complete | Sequence: F1 → (F2 ∥ F3 ∥ F4 ∥ F5 in parallel) → F6 → F7. The middle batch is four parallel `Task` calls in one message. |\n| **Part 10 — Execution Artefacts** | Ad copy, post copy, headlines, CTAs across channels — independent per channel | Dispatch one subagent per channel in parallel |\n| **Part 11 — AI Creative Instructions** | Visual asset briefs — independent per asset | Dispatch in parallel per asset |\n\n**Parts that MUST stay sequential** (have hard data dependencies):\n\n- Part 1 → Part 2 (intake feeds research scope)\n- Part 3 → Part 4 (Four Core Documents feed competitive\u002Fcustomer\u002Fmarket analysis)\n- Part 5 → Part 6 (Client Validation drives which docs need v2 re-runs)\n- Part 7 → Part 8 (prep docs feed the Growth Plan)\n- Part 8 → Part 9 (Growth Plan drives channel fan-out)\n\n**Cross-cutting rules:**\n\n1. Never dispatch parallel agents that need to write to the same file simultaneously — chunk by output file.\n2. Each parallel subagent gets the engagement slug and the LIF path so it can read shared context, but writes ONLY to its own numbered per-part subdirectory (01-… 12-…).\n3. **Subagents never mutate engagement state.** A subagent must NOT call `lif-log-change`, `mark-part-completed`, `bump-version`, or any other `engagement-state.py` write, and must NOT touch `_engagement.json` or `living-instruction-file.md`. Those are unlocked read-modify-write files; concurrent writers lose updates. Each subagent returns its output as per-part files only. After a parallel batch completes, the **orchestrator alone** applies state mutations — one `lif-log-change` per batch, plus `mark-part-completed` \u002F `bump-version` as needed — and then re-reads the LIF before the next step.\n4. If a parallel batch fails partway, the failed subagent's outputs are NOT auto-rolled-back — re-dispatch only the failed ones; the successful peers stay valid.\n\nFor multi-dimensional commands outside the 12-part flow (e.g. `\u002Fdigital-marketing-pro:competitor-analysis`, `\u002Fdigital-marketing-pro:seo-audit`, `\u002Fdigital-marketing-pro:content-engine`), the same pattern applies — dispatch independent dimensions in parallel via multiple `Task` calls in a single message.\n\n## Running an engagement in a single conversation\n\nA large-context model can hold much of an engagement — intake, external research, the Four Core Documents (61 steps), competitive\u002Fcustomer\u002Fmarket analysis, Client Validation, selective v2 re-runs, preparation docs, Growth Plan + Yearly Planner, channel fan-out, execution artefacts, creative briefs, and the continuous-improvement loop — within one working session (a full engagement typically produces 50–60 canonical documents).\n\n**The checkpoint + persistence pattern is still the default — always.** Even when everything fits in one conversation:\n\n- `engagement-state.py` + `checkpoint-manager.py` remain the system of record: audit trail, cross-conversation resume, and multi-user \u002F multi-day continuity all depend on persisted state.\n- Do NOT skip LIF updates or state writes on the assumption that \"it's all in context.\" An interruption still loses in-memory work, and a teammate resuming the engagement reads persisted state — not your conversation.\n- The only single-conversation convenience is that you re-read fewer files mid-session because they are already in context. It does not remove the need to persist, checkpoint, and update the LIF.\n\n## Quality Discipline\n\n1. **Never hand-edit `_engagement.json`.** Always go through `engagement-state.py`.\n2. **Never delete v1.** When v2 is produced, both stay.\n3. **Always update the LIF when source docs change.** Use `lif-log-change` — it appends to the change-log section of `living-instruction-file.md` and refreshes the header date, so the LIF never goes stale.\n4. **Always cite source per fact.** Stone facts cite the validation source; Opinion hypotheses cite the client's evidence.\n5. **Never auto-advance parts.** The user confirms part completion explicitly.\n6. **Always parallelize independent work.** When a Part has 2+ independent sub-tasks (see Parallel Dispatch above), dispatch them in a single message with multiple Task calls. Sequential dispatch of independent work wastes wall-clock time and API turns.\n7. **Never auto-execute v2 re-runs without user approval.** Show the plan, get approval, then run. Cap: 2 v2 re-run rounds per part without an explicit user override (stored in state).\n\n## Examples\n\n### Example: Starting a new engagement\n\n```\nUser: I want to start a new engagement for Acme Corp this quarter.\n\nSkill response:\n1. Verify Acme has a brand profile (slug: acme-corp). Found.\n2. Initialise engagement: engagement-state.py init --brand acme-corp --id 2026-q2\n3. Engagement directory created at ~\u002F.claude-marketing\u002Fbrands\u002Facme-corp\u002Fengagements\u002F2026-q2\u002F\n4. Walk Part 1 intake. Begin with Stone questions:\n   \"Let's start with what we know for certain. What year was Acme founded? Where is it headquartered? How many employees today?\"\n5. (User answers; each fact captured via add-stone-fact)\n6. Move to Opinion questions:\n   \"Now let's capture what you believe but haven't formally validated. How would you describe Acme's positioning in the market today?\"\n7. (User answers; each hypothesis captured via add-opinion)\n8. Mark Part 1 completed. Advise: \"Part 1 is complete. The unbiased External Research (Part 2) is next — recommend running it now or scheduling it for the next session.\"\n```\n\n### Example: Mid-engagement status check\n\n```\nUser: Where are we on the Beta Industries engagement?\n\nSkill response:\n- Reads engagement-state.py status --brand beta-industries --id 2026-h1\n- Shows:\n  Engagement: beta-industries \u002F 2026-h1\n  Started: 2026-04-01 (32 days in)\n  Current part: Part 9 — Channel Strategy Fan-out (in progress)\n  Completed parts: 1, 2, 3, 4, 5, 6, 7, 8\n  Pending parts: 9 (in progress), 10, 11, 12\n  Open re-run decisions: 1 (Part 6 v2 re-runs in progress)\n  Open items requiring resolution: 2 (LinkedIn Ads CPL forecast pending; Q3 budget envelope pending)\n  LIF last updated: 2 days ago\n  Next required action: complete remaining Part 9 channel docs (5 of 9 complete)\n```\n\n### Example: Applying the Decision Matrix\n\n```\nUser: We just finished Part 5 client validation. The client rejected our positioning recommendation and added two new competitors. What re-runs do we need?\n\nSkill response:\n1. Categorise the changes:\n   - Two new competitors → trigger \"competitors_changed\"\n   - Positioning rejected → trigger \"positioning_changed\"\n2. Run engagement-state.py decision-matrix --triggers \"competitors_changed,positioning_changed\"\n3. Output the triggered re-runs:\n   Triggered re-runs: 3.1, 3.2, 3.3, 3.4, 4.1, 4.2\n   (competitors_changed triggers all four Core Docs + 4.1 + 4.2; positioning_changed adds 3.3 which is already in scope)\n4. Estimate token cost: ~85K tokens for the full re-run set\n5. Ask: \"Approve all? Modify? Skip any?\"\n6. On user approval, mark Part 6 started and invoke four-core-documents skill with view=v2 for each triggered doc\n```\n\n### Example: Update-Back correction in flight\n\n```\nUser: We've been live for two months. Q1 channel data shows our actual CAC for Segment X is INR 4,800 — but Core Doc 3.1 v2.0 had it estimated at INR 3,000. Need to update.\n\nSkill response:\n1. Validate the correction: 60+ days of data, statistically meaningful sample. ✓\n2. Read current 3.1 v2.0\n3. Bump version: engagement-state.py bump-version --doc 3.1 --reason \"Segment X CAC corrected from INR 3,000 to INR 4,800 based on Q1 channel data\"\n4. Save 3.1 v2.1 with header note: \"v2.0 → v2.1 changes: Section 4 (Unit Economics) — Segment X CAC corrected. LTV:CAC drops from 4.0 to 2.5; below 3.0 health threshold.\"\n5. lif-log-change --section \"Recent Corrections\" --summary \"Segment X CAC corrected to INR 4,800. Channel mix for Segment X under review.\"\n6. Identify downstream review needed: Core Doc 3.4 (DMFlow) channel allocation for Segment X may need revision\n7. Add to engagement review queue\n```\n\n## Related skills\n\n- `four-core-documents` — produces Part 3 deliverables\n- `client-validation-document` — produces Part 5 deliverable\n- `growth-plan` + `yearly-planner` — produce Part 8 deliverables\n- `continuous-improvement-loop` — handles Part 12\n\n## Related references\n\n- [engagement-flow-methodology.md](..\u002Fcontext-engine\u002Fengagement-flow-methodology.md)\n- [four-core-documents-spec.md](..\u002Fcontext-engine\u002Ffour-core-documents-spec.md)\n- [decision-matrix-rerun.md](..\u002Fcontext-engine\u002Fdecision-matrix-rerun.md)\n- [update-back-rule.md](..\u002Fcontext-engine\u002Fupdate-back-rule.md)\n- [living-instruction-file-spec.md](..\u002Fcontext-engine\u002Fliving-instruction-file-spec.md)\n- [stone-vs-opinion.md](..\u002Fcontext-engine\u002Fstone-vs-opinion.md)\n- [two-views-model.md](..\u002Fcontext-engine\u002Ftwo-views-model.md)\n","# \u002Fdigital-marketing-pro:engagement-workflow — 12-Part Engagement Orchestrator\n\nThis skill orchestrates the full marketing engagement using the 12-Part sequential methodology. Every brand engagement runs through the same 12 parts in sequence, producing a canonical set of files at each stage.\n\n## Context efficiency\n\nHeavy skill. **Grep before Read** any referenced file, then `Read` only matched ranges with `offset` + `limit`. List the brand's workspace at `~\u002F.claude-marketing\u002Fbrands\u002F{slug}\u002F` (or `$CLAUDE_PLUGIN_DATA\u002Fdigital-marketing-pro\u002Fbrands\u002F{slug}\u002F` when that env var is set) before opening files. On re-invocation mid-session, skip files already in context.\n\nRead these references before producing output:\n- [engagement-flow-methodology.md](..\u002Fcontext-engine\u002Fengagement-flow-methodology.md) — the full 12-Part flow\n- [two-views-model.md](..\u002Fcontext-engine\u002Ftwo-views-model.md) — v1 \u002F v2 architecture\n- [stone-vs-opinion.md](..\u002Fcontext-engine\u002Fstone-vs-opinion.md) — confidence tagging\n- [decision-matrix-rerun.md](..\u002Fcontext-engine\u002Fdecision-matrix-rerun.md) — when to re-run what\n- [update-back-rule.md](..\u002Fcontext-engine\u002Fupdate-back-rule.md) — versioning protocol\n- [living-instruction-file-spec.md](..\u002Fcontext-engine\u002Fliving-instruction-file-spec.md) — LIF schema\n\n## Operating Mode\n\nThis skill is invoked via the `\u002Fdigital-marketing-pro:engagement` command family. The command is a thin router — **this skill is the single source of truth** for the engagement lifecycle, the checkpoint protocol, and the per-part production contract. Each subcommand maps to a specific lifecycle action. The skill calls `engagement-state.py` for persistence via:\n\n```\npython \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fengagement-state.py\" \u003Csubcommand> ...\n```\n\nYou should never hand-edit `_engagement.json` — always go through `engagement-state.py`.\n\n## Checkpointing & Resume (single source of truth)\n\nEvery long engagement run is resumable. The checkpoint protocol is: **init a run → save each part as it completes → finalize → publish to the visible output folder.** This lets an interrupted run (context exhaustion, user cancel, machine sleep) resume from the next un-checkpointed part instead of restarting from Part 1.\n\n**1. On `start`, after the brand pre-condition passes, open a checkpoint run and link it to engagement state:**\n\n```bash\npython \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fcheckpoint-manager.py\" init \\\n    --brand \"{brand_slug}\" --workflow engagement --topic \"{engagement_id}\"\n\n# Record the returned run_id into _engagement.json so resume can find it:\npython \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fengagement-state.py\" set-checkpoint-run \\\n    --brand \"{brand_slug}\" --id \"{engagement_id}\" --run-id \"{run_id}\"\n```\n\n`set-checkpoint-run` stores the run_id in `_engagement.json`, making the resume linkage real (previously the run_id was never persisted).\n\n**2. After each part completes and passes its quality gate, the orchestrator saves that part's output:**\n\n```bash\npython \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fcheckpoint-manager.py\" save \\\n    --brand \"{brand}\" --run-id \"{run_id}\" \\\n    --step {part_number} --content-file \"{path_to_that_part_deliverable}\" --extension md\n```\n\nPass the **actual deliverable path for that part** (e.g. Part 3 saves the Four Core Documents path; Part 8 saves the Growth Plan path) — never a placeholder for a different part.\n\n**3. Before saving Part 5 (Client Validation) and Part 8 (Growth Plan) deliverables, run the full quality gate:**\n\n```bash\n# BLOCKING gate — Part 5 and Part 8 deliverables cannot be checkpointed until this passes\n\u002Fdigital-marketing-pro:check \"{path_to_deliverable}\" --full --brand {brand}\n```\n\nIf `\u002Fdigital-marketing-pro:check --full` returns BLOCKED, fix the CRITICAL issues before checkpointing the part.\n\n**4. After the final part, publish every artifact to the user-visible folder and finalize:**\n\n```bash\npython \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Foutput-publisher.py\" publish-run \\\n    --brand \"{brand}\" --run-id \"{run_id}\"\n\npython \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fcheckpoint-manager.py\" finalize \\\n    --brand \"{brand}\" --run-id \"{run_id}\" --status completed\n```\n\nThen point the user at the visible output folder via `\u002Fdigital-marketing-pro:output-folder {brand}`.\n\nTo resume an interrupted run, use `\u002Fdigital-marketing-pro:resume` — it reloads every saved part and continues from the next un-checkpointed part.\n\n## State validation & rework caps\n\n- **Validate a part's outputs against the manifest** before marking it complete:\n  ```bash\n  python \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fengagement-state.py\" validate-part \\\n      --brand \"{brand}\" --id \"{id}\" --part {N}\n  ```\n  This diffs the actual files on disk against the `PART_DEFINITIONS` manifest and flags missing deliverables. Use it in `file-tree` and before `next`.\n\n- **Repair a partially-initialised engagement directory** (instead of crashing on a non-empty dir):\n  ```bash\n  python \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fengagement-state.py\" init --repair \\\n      --brand \"{brand}\" --id \"{id}\"\n  ```\n  `--repair` completes the canonical directory tree and state file on a dir that holds only partial state.\n\n- **v2 re-run cap:** a maximum of **2 v2 re-run rounds per part** is allowed without explicit user override. The round count is stored in `_engagement.json`. If a part would exceed 2 rounds, stop and ask the user to explicitly approve further re-runs (records the override in state). This prevents unbounded re-run loops.\n\n## Subcommands\n\n### `\u002Fdigital-marketing-pro:engagement start \u003Cbrand-slug> \u003Cengagement-id>`\n\n**Purpose:** Initialise a new engagement.\n\n**Steps:**\n\n1. Validate that the brand profile exists at `~\u002F.claude-marketing\u002Fbrands\u002F{brand-slug}\u002Fprofile.json`. If not, instruct the user to run `\u002Fdigital-marketing-pro:brand-setup` first.\n2. Run `python ${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fengagement-state.py init --brand {brand-slug} --id {engagement-id}`.\n3. Confirm the directory tree was created and report the next required action (Part 1 intake).\n4. Walk the user through Part 1 Stone vs Opinion intake by asking the questions one batch at a time.\n\n**Part 1 intake questions (ask in this order):**\n\n**Stone — what the client knows for certain:**\n\n1. Company basics: founded year, employee count, headquarters location, geographic operations\n2. Business model: revenue streams, pricing tiers, primary product\u002Fservice categories\n3. Current marketing: channels currently active, monthly marketing spend, current measurable KPIs\n4. Tech stack: CRM, email service provider, analytics setup, ad accounts\n5. Customer base scale: customer count, biggest named customer, average order value if known\n\nFor each Stone fact, capture:\n- The fact itself\n- Source (how the user knows \u002F what document confirmed it)\n\nSave each via:\n```\npython ${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fengagement-state.py add-stone-fact --brand {slug} --id {id} --fact-json '{\"category\":\"...\",\"fact\":\"...\",\"source\":\"...\"}'\n```\n\n**Opinion — what the client believes:**\n\n1. Brand positioning: how does the client describe their position in the market?\n2. Customer base: who do they think their customers are? Why do they buy?\n3. Competitors: who do they consider their main competitors?\n4. Growth opportunities: where do they think the biggest opportunity is?\n5. What is working: what marketing activity does the client believe is working?\n6. What is not working: what does the client believe is not working?\n\nFor each Opinion, capture:\n- The hypothesis\n- Client's evidence for it (could be intuition, anecdote, partial data)\n- Research question — what would the unbiased research need to verify or refute?\n\nSave each via:\n```\npython ${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fengagement-state.py add-opinion --brand {slug} --id {id} --hypothesis-json '{\"category\":\"...\",\"hypothesis\":\"...\",\"client_evidence\":\"...\",\"research_question\":\"...\"}'\n```\n\n**On completion of Part 1:** mark Part 1 as completed via `mark-part-completed --part 1`, advise the user to proceed to Part 2 (External Research).\n\n### `\u002Fdigital-marketing-pro:engagement next [brand] [id]`\n\n**Purpose:** Advance to the next part.\n\n**Steps:**\n\n1. Read engagement status via `engagement-state.py status`\n2. Identify the current part and next not-yet-completed part\n3. Confirm with the user that the current part is genuinely complete (do not auto-advance — ask)\n4. On confirmation, mark current as completed, advance current_part pointer\n5. Brief the user on what the new part requires\n\n### `\u002Fdigital-marketing-pro:engagement status [brand] [id]`\n\n**Purpose:** Show engagement status.\n\n**Steps:**\n\n1. Run `engagement-state.py status` — get the full state\n2. Read the Living Project Instruction File header\n3. Format a human-readable summary:\n   - Engagement: brand + id + start date\n   - Current part: part name + days in\n   - Completed parts: list\n   - Pending parts: list\n   - Open re-run decisions: count\n   - LIF last updated: date\n4. If the engagement has open items needing resolution, list them\n\n### `\u002Fdigital-marketing-pro:engagement file-tree [brand] [id]`\n\n**Purpose:** Show the engagement directory file tree.\n\n**Steps:**\n\n1. Run `engagement-state.py file-tree`\n2. Format as an indented tree\n3. Highlight files that are missing per the canonical structure. Use `engagement-state.py validate-part --part {N}` to diff each completed part's actual files against the `PART_DEFINITIONS` manifest (e.g., if Part 3 is marked completed but `3.1-business-and-sbu-analysis.md` is missing, `validate-part` flags it deterministically instead of eyeballing).\n\n### `\u002Fdigital-marketing-pro:engagement validate [brand] [id]`\n\n**Purpose:** Run the Part 5 Client Validation flow.\n\n**Pre-condition:** Parts 2, 3, 4 must be completed.\n\n**Steps:**\n\n1. Verify pre-conditions (Parts 2, 3, 4 completed)\n2. Invoke the `client-validation-document` skill — it produces the Part 5 deliverable: a structured document presenting each finding from v1 with ACCEPT\u002FREJECT\u002FEDIT\u002FDEFER options\n3. **Run the full quality gate on the Part 5 deliverable before it goes to the client:** `\u002Fdigital-marketing-pro:check \"{part5_path}\" --full --brand {brand}`. If it returns BLOCKED, fix the CRITICAL issues first (this gate is mandatory before Part 5 and Part 8 deliverables).\n4. After the user reviews and provides decisions, parse them into a triggers list per the Decision Matrix categories\n5. Run `engagement-state.py decision-matrix --triggers \"{comma-separated}\"` to compute the v2 re-run plan\n6. Present the re-run plan to the user\n7. Mark Part 5 completed; on user approval of the re-run plan, advance to Part 6\n\n### `\u002Fdigital-marketing-pro:engagement re-run-decision [brand] [id]`\n\n**Purpose:** Apply the Decision Matrix to compute v2 re-runs.\n\n**Steps:**\n\n1. Read the Part 5 Client Validation Document\n2. Categorise rejected\u002Fedited findings into Decision Matrix triggers\n3. Show the triggers and the computed re-runs\n4. Estimate the cost (rough token count) of each re-run\n5. Await user approval — they can accept, modify (skip some, add others), or reject\n6. Record the executed plan via `engagement-state.py record-rerun-execution`\n\n### `\u002Fdigital-marketing-pro:engagement update-back [brand] [id] --doc \u003Cdoc-id> --reason \u003Creason>`\n\n**Purpose:** Apply the Update-Back Rule to bump a source document version after Part 7+.\n\n**Pre-condition:** The user has already drafted the corrected document content.\n\n**Steps:**\n\n1. Read the current version of the doc\n2. Confirm the correction with the user (validation step per the Update-Back Rule)\n3. Bump the version via `engagement-state.py bump-version --doc {id} --reason \"{reason}\"`\n4. Save the new version file with a header noting v(prev) → v(new) changes\n5. Update the Living Project Instruction File via `lif-log-change` — it now appends the change to `living-instruction-file.md` and refreshes the header date, so the LIF reflects the correction immediately\n6. Identify downstream documents that may need review and add to the engagement's review queue\n\n### `\u002Fdigital-marketing-pro:engagement lif-show [brand] [id]`\n\n**Purpose:** Display the Living Project Instruction File.\n\n**Steps:** Run `engagement-state.py lif-show` and format the markdown output for readability.\n\n### `\u002Fdigital-marketing-pro:engagement list-engagements [brand]`\n\n**Purpose:** List all engagements (optionally filtered by brand).\n\n**Steps:** Run `engagement-state.py list-engagements --brand {slug}` and format as a table.\n\n### Production shorthands\n\nThe command family also exposes four production shorthands that route straight to the part-producing skills (documented in *Per-Part Production Targets* below). These match the command surface one-to-one:\n\n- `\u002Fdigital-marketing-pro:engagement four-core \u003Cbrand> \u003Cid> [--doc 3.X] [--view v2] [--combined]` — Part 3, invokes the `four-core-documents` skill\n- `\u002Fdigital-marketing-pro:engagement growth-plan \u003Cbrand> \u003Cid>` — Part 8, invokes the `growth-plan` skill\n- `\u002Fdigital-marketing-pro:engagement yearly-planner \u003Cbrand> \u003Cid>` — Part 8 companion, invokes the `yearly-planner` skill\n- `\u002Fdigital-marketing-pro:engagement loop \u003Cbrand> \u003Cid>` — Part 12, invokes the `continuous-improvement-loop` skill\n\n## Per-Part Production Targets\n\nEach part is produced by real, existing agents and skills. This orchestrator dispatches to the targets below — there are **no** wrapper skills named `external-research` \u002F `preparation-documents` \u002F `channel-strategy-fanout` \u002F `execution-artefacts` \u002F `ai-creative-instructions`; those never existed. Use the exact targets named here:\n\n| Part | Real target(s) |\n|------|----------------|\n| 1 | (this skill — intake walked here directly) |\n| 2 | agents `market-intelligence` + `competitive-intel`; skill `audience-intelligence` (invoke as a skill); reference `compliance-rules.md` (load as context) |\n| 3 | skill `four-core-documents` (produces 3.1, 3.2, 3.3, 3.4) |\n| 4 | command `competitor-analysis` + skills `audience-intelligence` + agent `market-intelligence` |\n| 5 | skill `client-validation-document` |\n| 6 | re-runs invoke skill `four-core-documents` with `--view v2` |\n| 7 | skills `content-engine` + `campaign-orchestrator` + `analytics-insights` |\n| 8 | skills `growth-plan` + `yearly-planner` |\n| 9 | per-channel skills — `paid-advertising`, `aeo-geo`, `social-strategy`, `seo-plan`, `email-sequence` (one per channel family) |\n| 10 | skill `content-engine` (execution \u002F output mode) |\n| 11 | skills `content-engine` + `ad-creative` + `video-script` (creative briefs); asset rendering happens in your own creative tooling (design team, AI image\u002Fvideo tools, or a connected design platform), then finished assets are signed via `c2pa-metadata` |\n| 12 | skill `continuous-improvement-loop` |\n\n## Parallel Dispatch\n\nSeveral parts of the engagement contain **independent sub-tasks** that should be dispatched **in parallel via multiple `Task` tool calls in a single message** — not sequentially. Dispatching independent sub-tasks concurrently is substantially faster than running them one after another; actual time varies by engagement depth, model, and rate limits. Keep concurrent subagents to a handful (roughly 3–8) — past that you queue against API rate limits and the win drops; under 3 there is nothing to parallelize.\n\n**Cost note:** total token usage is broadly similar (you're doing the same work) but billed-per-turn input costs trend up slightly because each parallel subagent re-loads its context.\n\n**Parts that benefit from parallel dispatch:**\n\n| Part | Parallel-eligible work | How to dispatch |\n|---|---|---|\n| **Part 2 — External Research** | Market sizing, competitor landscape, customer signals, regulatory landscape — none depend on each other | Dispatch the `market-intelligence` agent and the `competitive-intel` agent as parallel `Task` calls; invoke `audience-intelligence` as a skill; load `compliance-rules.md` (a reference file) as context — not as a subagent |\n| **Part 4 — Competitive + Customer + Market** | Four documents (4.1, 4.2, 4.3, 4.4) are independent — they reference Part 2 only | Dispatch all four in a single message with the four respective subagents |\n| **Part 9 — Channel Strategy Fan-out** | Up to 17 channel docs in 7 families. Families 2 (Paid platforms), 3 (Organic & Influencer), 4 (Marketplace & CRM), 5 (Content\u002FATL\u002FBTL\u002FPR) are independent after Families 1 (Search & Campaign) and 6 (Web + Measurement) complete | Sequence: F1 → (F2 ∥ F3 ∥ F4 ∥ F5 in parallel) → F6 → F7. The middle batch is four parallel `Task` calls in one message. |\n| **Part 10 — Execution Artefacts** | Ad copy, post copy, headlines, CTAs across channels — independent per channel | Dispatch one subagent per channel in parallel |\n| **Part 11 — AI Creative Instructions** | Visual asset briefs — independent per asset | Dispatch in parallel per asset |\n\n**Parts that MUST stay sequential** (have hard data dependencies):\n\n- Part 1 → Part 2 (intake feeds research scope)\n- Part 3 → Part 4 (Four Core Documents feed competitive\u002Fcustomer\u002Fmarket analysis)\n- Part 5 → Part 6 (Client Validation drives which docs need v2 re-runs)\n- Part 7 → Part 8 (prep docs feed the Growth Plan)\n- Part 8 → Part 9 (Growth Plan drives channel fan-out)\n\n**Cross-cutting rules:**\n\n1. Never dispatch parallel agents that need to write to the same file simultaneously — chunk by output file.\n2. Each parallel subagent gets the engagement slug and the LIF path so it can read shared context, but writes ONLY to its own numbered per-part subdirectory (01-… 12-…).\n3. **Subagents never mutate engagement state.** A subagent must NOT call `lif-log-change`, `mark-part-completed`, `bump-version`, or any other `engagement-state.py` write, and must NOT touch `_engagement.json` or `living-instruction-file.md`. Those are unlocked read-modify-write files; concurrent writers lose updates. Each subagent returns its output as per-part files only. After a parallel batch completes, the **orchestrator alone** applies state mutations — one `lif-log-change` per batch, plus `mark-part-completed` \u002F `bump-version` as needed — and then re-reads the LIF before the next step.\n4. If a parallel batch fails partway, the failed subagent's outputs are NOT auto-rolled-back — re-dispatch only the failed ones; the successful peers stay valid.\n\nFor multi-dimensional commands outside the 12-part flow (e.g. `\u002Fdigital-marketing-pro:competitor-analysis`, `\u002Fdigital-marketing-pro:seo-audit`, `\u002Fdigital-marketing-pro:content-engine`), the same pattern applies — dispatch independent dimensions in parallel via multiple `Task` calls in a single message.\n\n## Running an engagement in a single conversation\n\nA large-context model can hold much of an engagement — intake, external research, the Four Core Documents (61 steps), competitive\u002Fcustomer\u002Fmarket analysis, Client Validation, selective v2 re-runs, preparation docs, Growth Plan + Yearly Planner, channel fan-out, execution artefacts, creative briefs, and the continuous-improvement loop — within one working session (a full engagement typically produces 50–60 canonical documents).\n\n**The checkpoint + persistence pattern is still the default — always.** Even when everything fits in one conversation:\n\n- `engagement-state.py` + `checkpoint-manager.py` remain the system of record: audit trail, cross-conversation resume, and multi-user \u002F multi-day continuity all depend on persisted state.\n- Do NOT skip LIF updates or state writes on the assumption that \"it's all in context.\" An interruption still loses in-memory work, and a teammate resuming the engagement reads persisted state — not your conversation.\n- The only single-conversation convenience is that you re-read fewer files mid-session because they are already in context. It does not remove the need to persist, checkpoint, and update the LIF.\n\n## Quality Discipline\n\n1. **Never hand-edit `_engagement.json`.** Always go through `engagement-state.py`.\n2. **Never delete v1.** When v2 is produced, both stay.\n3. **Always update the LIF when source docs change.** Use `lif-log-change` — it appends to the change-log section of `living-instruction-file.md` and refreshes the header date, so the LIF never goes stale.\n4. **Always cite source per fact.** Stone facts cite the validation source; Opinion hypotheses cite the client's evidence.\n5. **Never auto-advance parts.** The user confirms part completion explicitly.\n6. **Always parallelize independent work.** When a Part has 2+ independent sub-tasks (see Parallel Dispatch above), dispatch them in a single message with multiple Task calls. Sequential dispatch of independent work wastes wall-clock time and API turns.\n7. **Never auto-execute v2 re-runs without user approval.** Show the plan, get approval, then run. Cap: 2 v2 re-run rounds per part without an explicit user override (stored in state).\n\n## Examples\n\n### Example: Starting a new engagement\n\n```\nUser: I want to start a new engagement for Acme Corp this quarter.\n\nSkill response:\n1. Verify Acme has a brand profile (slug: acme-corp). Found.\n2. Initialise engagement: engagement-state.py init --brand acme-corp --id 2026-q2\n3. Engagement directory created at ~\u002F.claude-marketing\u002Fbrands\u002Facme-corp\u002Fengagements\u002F2026-q2\u002F\n4. Walk Part 1 intake. Begin with Stone questions:\n   \"Let's start with what we know for certain. What year was Acme founded? Where is it headquartered? How many employees today?\"\n5. (User answers; each fact captured via add-stone-fact)\n6. Move to Opinion questions:\n   \"Now let's capture what you believe but haven't formally validated. How would you describe Acme's positioning in the market today?\"\n7. (User answers; each hypothesis captured via add-opinion)\n8. Mark Part 1 completed. Advise: \"Part 1 is complete. The unbiased External Research (Part 2) is next — recommend running it now or scheduling it for the next session.\"\n```\n\n### Example: Mid-engagement status check\n\n```\nUser: Where are we on the Beta Industries engagement?\n\nSkill response:\n- Reads engagement-state.py status --brand beta-industries --id 2026-h1\n- Shows:\n  Engagement: beta-industries \u002F 2026-h1\n  Started: 2026-04-01 (32 days in)\n  Current part: Part 9 — Channel Strategy Fan-out (in progress)\n  Completed parts: 1, 2, 3, 4, 5, 6, 7, 8\n  Pending parts: 9 (in progress), 10, 11, 12\n  Open re-run decisions: 1 (Part 6 v2 re-runs in progress)\n  Open items requiring resolution: 2 (LinkedIn Ads CPL forecast pending; Q3 budget envelope pending)\n  LIF last updated: 2 days ago\n  Next required action: complete remaining Part 9 channel docs (5 of 9 complete)\n```\n\n### Example: Applying the Decision Matrix\n\n```\nUser: We just finished Part 5 client validation. The client rejected our positioning recommendation and added two new competitors. What re-runs do we need?\n\nSkill response:\n1. Categorise the changes:\n   - Two new competitors → trigger \"competitors_changed\"\n   - Positioning rejected → trigger \"positioning_changed\"\n2. Run engagement-state.py decision-matrix --triggers \"competitors_changed,positioning_changed\"\n3. Output the triggered re-runs:\n   Triggered re-runs: 3.1, 3.2, 3.3, 3.4, 4.1, 4.2\n   (competitors_changed triggers all four Core Docs + 4.1 + 4.2; positioning_changed adds 3.3 which is already in scope)\n4. Estimate token cost: ~85K tokens for the full re-run set\n5. Ask: \"Approve all? Modify? Skip any?\"\n6. On user approval, mark Part 6 started and invoke four-core-documents skill with view=v2 for each triggered doc\n```\n\n### Example: Update-Back correction in flight\n\n```\nUser: We've been live for two months. Q1 channel data shows our actual CAC for Segment X is INR 4,800 — but Core Doc 3.1 v2.0 had it estimated at INR 3,000. Need to update.\n\nSkill response:\n1. Validate the correction: 60+ days of data, statistically meaningful sample. ✓\n2. Read current 3.1 v2.0\n3. Bump version: engagement-state.py bump-version --doc 3.1 --reason \"Segment X CAC corrected from INR 3,000 to INR 4,800 based on Q1 channel data\"\n4. Save 3.1 v2.1 with header note: \"v2.0 → v2.1 changes: Section 4 (Unit Economics) — Segment X CAC corrected. LTV:CAC drops from 4.0 to 2.5; below 3.0 health threshold.\"\n5. lif-log-change --section \"Recent Corrections\" --summary \"Segment X CAC corrected to INR 4,800. Channel mix for Segment X under review.\"\n6. Identify downstream review needed: Core Doc 3.4 (DMFlow) channel allocation for Segment X may need revision\n7. Add to engagement review queue\n```\n\n## Related skills\n\n- `four-core-documents` — produces Part 3 deliverables\n- `client-validation-document` — produces Part 5 deliverable\n- `growth-plan` + `yearly-planner` — produce Part 8 deliverables\n- `continuous-improvement-loop` — handles Part 12\n\n## Related references\n\n- [engagement-flow-methodology.md](..\u002Fcontext-engine\u002Fengagement-flow-methodology.md)\n- [four-core-documents-spec.md](..\u002Fcontext-engine\u002Ffour-core-documents-spec.md)\n- [decision-matrix-rerun.md](..\u002Fcontext-engine\u002Fdecision-matrix-rerun.md)\n- [update-back-rule.md](..\u002Fcontext-engine\u002Fupdate-back-rule.md)\n- [living-instruction-file-spec.md](..\u002Fcontext-engine\u002Fliving-instruction-file-spec.md)\n- [stone-vs-opinion.md](..\u002Fcontext-engine\u002Fstone-vs-opinion.md)\n- [two-views-model.md](..\u002Fcontext-engine\u002Ftwo-views-model.md)\n",{"name":5,"triggers":24,"description":7,"allowed-tools":33,"user-invocable":34,"engagement-part":35,"view-preference":36},[25,26,27,28,29,30,31,32],"start a new engagement","run the 12-part methodology","advance engagement to next part","show engagement status","apply the decision matrix","re-run v2 documents","mark engagement part complete","what part of the engagement are we on","Read Write Edit Bash Glob Grep",true,"orchestrator","both","main","4fd50b1281e8cd978866b6b4556242605cabf708","2_KNOWLEDGE\u002Fframeworks\u002Fdigital-marketing-pro\u002Fskills\u002Fengagement-workflow",1,"https:\u002F\u002Fgithub.com\u002FLongLeo287\u002FSEOSONA-OS","https:\u002F\u002Fseosona-os.vercel.app","https:\u002F\u002Fgithub.com\u002FLongLeo287","https:\u002F\u002Fwww.facebook.com\u002FResourceRookie2023","2026-10-09 15:14:13",{"source":8,"skillName":5,"commands":47,"prompt":57,"zipAvailable":34},[48,51,54],{"tool":49,"command":50},"npx","npx skills add LongLeo287\u002FSEOSONA-OS --skill engagement-workflow",{"tool":52,"command":53},"bunx","bunx skills add LongLeo287\u002FSEOSONA-OS --skill engagement-workflow",{"tool":55,"command":56},"pnpm","pnpm dlx skills add LongLeo287\u002FSEOSONA-OS --skill engagement-workflow","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FLongLeo287\u002FSEOSONA-OS\u002Fblob\u002Fmain\u002F2_KNOWLEDGE\u002Fframeworks\u002Fdigital-marketing-pro\u002Fskills\u002Fengagement-workflow\u002FSKILL.md",[],[60,67,74,81,88,95],{"skillNo":61,"name":62,"nameZh":6,"description":63,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":64,"categories":65,"keywords":66,"publishedAt":19},"01M75F0T3V3W7JQQ05X26NNN2T","web_crawler_katana","High-speed web crawling using Katana patterns for deep site mapping, URL discovery, and content extraction.","https:\u002F\u002Fgithub.com\u002FLongLeo287\u002FSEOSONA-OS\u002Fblob\u002Fmain\u002F2_KNOWLEDGE\u002Fskills\u002Fseo_technical\u002Fweb_crawler_katana\u002FSKILL.md",[],[],{"skillNo":68,"name":69,"nameZh":6,"description":70,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"publishedAt":19},"01M75F0Q2QJXDGAFKC7D84GT88","server_log_analyzer","Analyzes Nginx\u002FApache server logs to identify Googlebot crawl patterns, errors, and performance bottlenecks.","https:\u002F\u002Fgithub.com\u002FLongLeo287\u002FSEOSONA-OS\u002Fblob\u002Fmain\u002F2_KNOWLEDGE\u002Fskills\u002Fseo_technical\u002Fserver_log_analyzer\u002FSKILL.md",[],[],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":19},"01M75F0KZP6D8R8XPN2YT35BAJ","schema_markup_generator","Auto-generates JSON-LD structured data for any page type (Article, Product, FAQ, LocalBusiness, etc.).","https:\u002F\u002Fgithub.com\u002FLongLeo287\u002FSEOSONA-OS\u002Fblob\u002Fmain\u002F2_KNOWLEDGE\u002Fskills\u002Fseo_technical\u002Fschema_markup_generator\u002FSKILL.md",[],[],{"skillNo":82,"name":83,"nameZh":6,"description":84,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":85,"categories":86,"keywords":87,"publishedAt":19},"01M75F0G3KX4Y86AXJXTKXVEVA","internal_linking_optimizer","Analyzes and optimizes the internal link graph of a website for SEO authority distribution.","https:\u002F\u002Fgithub.com\u002FLongLeo287\u002FSEOSONA-OS\u002Fblob\u002Fmain\u002F2_KNOWLEDGE\u002Fskills\u002Fseo_technical\u002Finternal_linking_optimizer\u002FSKILL.md",[],[],{"skillNo":89,"name":90,"nameZh":6,"description":91,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":92,"categories":93,"keywords":94,"publishedAt":19},"01M75F0CS1RYX4YPJC7N6RAAQJ","google_trends_analyzer","Detects trending topics and seasonal patterns using Google Trends data for content planning and keyword strategy.","https:\u002F\u002Fgithub.com\u002FLongLeo287\u002FSEOSONA-OS\u002Fblob\u002Fmain\u002F2_KNOWLEDGE\u002Fskills\u002Fseo_technical\u002Fgoogle_trends_analyzer\u002FSKILL.md",[],[],{"skillNo":96,"name":97,"nameZh":6,"description":98,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":99,"categories":100,"keywords":101,"publishedAt":19},"01M75F0ASMQ1GKPFX82D19SRTN","brand_context_analyzer","Tracks brand mentions, sentiment, and online reputation across web sources.","https:\u002F\u002Fgithub.com\u002FLongLeo287\u002FSEOSONA-OS\u002Fblob\u002Fmain\u002F2_KNOWLEDGE\u002Fskills\u002Fseo_technical\u002Fbrand_context_analyzer\u002FSKILL.md",[],[]]