Give it an objective, and it produces an entire slate of content planning.The human only reviews one table; the rest is mechanically derived.
This is a dispatcher. The map below tells you which layer each references/ file,load on demand, do not read it all at once。
Why it is different
The shape of most content tools is "give me a topic, and I will give you a draft." This set is the reverse:You give me a business objective, and I first tell you who this slate should target, what to target them with, and how many pieces, and only then write drafts.
Three ways it differs from others:
- The mix is derived, not copied from ratios. There is no empirical research supporting any content-piece-count ratio across funnel stages (80/20, 4-1-1, 70/20/10 are all assertions, and 70/20/10 is even a risk allocation miscited as a funnel allocation). So the numbers here come from yourcapacity, yourfunnel diagnosis, and yourrunway, not from a slogan.Capacity caps it, the objective shapes it, the diagnosis refines the shape, the format lands in place.
- The human only reviews the seed, not the drafts. A seed sheet of 8 to 20 entries is the only manual decision point. A content map of a dozen or so cells is mechanically derived from the seed and is reproducible.No one reviews those twenty-some cells one by one; scale should serve human judgment, not drown it.
- It will refuse. If follower scale is insufficient, it will not output a mix; it will instead output a reach-acquisition task list. If there is no real material on hand, it will not write drafts; it will instead output collection instructions."Not enough material, go get material first" is a legitimate, correct output.
When to use
- ✅ Planning an entire slate of content (a month, a launch, a runway)
- ✅ Content keeps going out but has no effect, and you want to know which segment is leaking
- ✅ You want to know how many videos and how many image-text pieces this slate should produce, instead of going by feel
- ❌ You just want to write one piece of copy → no need to run the whole pipeline
- ❌ Paid placement is the main channel → the mix logic is entirely different; this only covers organic
- ❌ No product yet, and no one has bought yet → that is a product-validation problem, not a content problem
Four modes
This skill has four entry paths.First determine which one it is, then proceed.
| Mode | When | What it does | Load |
|---|---|---|---|
| Plan | "Help me plan this month," "How many pieces should go out this slate" | The main line from objective to finished drafts, ending with revision backflow | modes/plan.md |
| Capture | "I picked up a few things customers said," "I have filled in those NEEDS-CAPTURE items" | Collect real material, sort scene_id, upgrade the barrier evidence level |
modes/capture.md |
| Log | Doorbell roll call, or "jot down the results" | Pure entry at 3 / 7 / 15 days, done in 30 seconds,no judgment made | modes/log.md |
| Review | "Let's do a retrospective," or after one cohort has run for 6 to 8 weeks | read the data, rerun diagnostics and ratios,produce proposals, not conclusions | modes/review.md |
Capture is the entire engine's only true evidence entry point. Everything that comes in from elsewhere is context.
If you have a second-brain vault
if you can see 99_Meta/structure-doctrine.md , then you have one. If you do, at the start read references/_vault.md: scan the business wing for brand context, read theoverlay(numbers the user has calibrated themselves take precedence over factory values), and know where things should live.
🛑 The first action of every mode is to get your bearings（_vault.md §1.5): a single glob to find content-engine-paths.md , this address book; all paths come from it. ⛔ No distinguishing between the first time and the hundredth time, ⛔ no guessing paths from memory: the folder names for the business wing and the brand arechosen by the user themselves, and cannot be guessed.
⚠️ What gets scanned back is context, not evidence. The eight-pillar foundation and brand strategy report read like evidence, even the quotation marks look like it, but in testing a locked 100KB strategy produced 0 confirmed customer barriers. So the first action when scanning the vault is a quote-provenance scan; anything unlabeled is treated as constructed.
It runs the same without a vault, following 00-intake.md to ask those six things. ⛔ A vault is a bonus, not a prerequisite.
There is one thing to do at install time
If the user already has the command-base skill installed,then when installing this skill you need to modify their already-installed copy of(not the product template; the template only affects future new installs), so that the doorbell recognizes content expiry.
⚠️ The product-side doorbell dogma:the bell rings only once, and the bell itself names what has expired; it does not toss a multiple-choice question at the owner(the product modes/maintenance.md 's "One doorbell, two files" is talking about exactly this). Sodo not add a second bell, just write "which pieces have hit 3 / 7 / 15 days" somewhere the existing doorbell can read, the bell still rings once, just names one more thing.
⛔ What you modify is the copy Claude Code actually loads（~/.claude/skills/<slug>-command-base/SKILL.md); the copy in the vault is only a replica under copy-install, modifying it will neither error nor take effect. After modifying, grep to read it back and confirm. ⛔ Do not write into the doorbell block itself; that block gets wholesale-overwritten on product version upgrades.
When running, go through these layers in order
| Layer | What it does | Who does it | Gear |
|---|---|---|---|
| L0 | Collect objective, deadline, capacity, platform and follower scale | Ask | 00-intake.md |
| L0.5 | Scale gate: if scale is insufficient, no ratios come out | Gate | 00-intake.md |
| L1 | Funnel diagnosis, produce a stage-pressure profile | Engine proposes | 01-diagnosis.md |
| L2 | Main-axis proposal, 2 to 3 cards | Human picks (decision point 1) | 02-mainline.md |
| L3 | Ratios: how many pieces per stage per format, how many platforms | Engine proposes, human may change numbers | 03-package-model.md |
| L4 | Seed sheet | Human review (decision point 2, the only one) | 05-seed-sheet.md |
| L5 | Content map, mechanically derived | Engine | 06-content-map.md |
| L6 | Pick one cell and write the draft | Engine | 07 → 08 → 09 → 10 |
Load on demand
| What does the user want | Which one to read |
|---|---|
| Start a new slate of planning / "Help me plan this month" | 00-intake.md |
| "Why is nobody watching my content" / Diagnose which segment it's stuck at | 01-diagnosis.md |
| Set the main axis / "What angle should I talk about" / reframe | 02-mainline.md |
| "How many pieces should this slate put out" / Mix / Which platform to post on | 03-package-model.md |
| Pick content types / "What content do I have to post" | 04-content-types.md |
| Produce a seed sheet / Review seeds | 05-seed-sheet.md |
| Push out a content map / How the content connects to each other | 06-content-map.md |
| Write one piece / How one piece grows | 07-body-structures.md |
| How to post on a given platform | 08-formats/(pick one by platform) |
| Thinking of a title / opening line / video hook | 09-hooks/(first read _tournament.md） |
| The draft reads like AI / Check before finalizing | 10-voice-anti-ai.md |
| "I have no material to write about" | 11-capture-loop.md |
| Run a whole slate in one go / Batch | 12-batch.md |
| There's a vault / Override values / Where things should live | references/_vault.md |
| Want to change a rule / Want to know why a given rule is this way | references/_RULINGS.md(The arbitration source of truth, ⛔ to change a rule first change the slate body, then come back and record the ruling and reasoning) |
| The numbers and things this user has calibrated and learned on their own | vault's 04_Methodology/Playbooks/Content-Engine/ |
A few rules that run through everything
These few rules show up repeatedly across many slates, because they really do apply everywhere:
- The cover-the-brand-name test. Cover the brand name, and if this piece of content could be posted by any peer in the field, it fails. If it's obvious at a glance that it's you, only then does it succeed.
- The free-teacher trap. Knowledge is cheap, AI generates it freely;opinions are what's valuable. Practical-tips content is the easiest to fall into this, and every piece must carry your own method and stance.
- Deep ≠ unique. Dig down into emotion, dig into identity, and dig to the end and you often dig up a universal truth that applies to everyone. That isthe most high-grade homogeneity, not insight. Differentiation comes from reframe (your perspective), not from digging deeper.
- One pitch inside Awareness, and the three stages collapse into one. When the product appears in awareness-stage content, the reader just scrolls away. There are no exceptions to this.
- Only real scarcity is allowed, manufactured urgency is not. Using manufactured urgency at the wrong stage isa category error, not a stylistic choice.
- Real memories have waste material, fabricated memories don't. If in a draft every single detail happens to serve the argument, with not one extra, then it's most likely fabricated.And that extra detail must come from the same source as the load-bearing details, otherwise you could casually make up an irrelevant detail and pass, and this rule would be written for nothing.
One sentence
Content is slow work.First there's real material, then there's an angle, then there's content. What this set of tools can do is give every step a basis; it can't go pick back up that sentence your customer said for you.