30-Day Multi-Platform Content System . Agency OS

What it does

Produces a complete content sprint in one run: a Research_Intelligence_Report.md, a Brand_Voice_Guide.md, a Platform_Optimization_Guide.md, a Content_Creation_SOP.md, a 30-day calendar, a master content library spreadsheet, and the content itself in nine consolidated files. Content is split across three awareness tiers (cold, warm, hot) and three formats (video scripts, image posts, carousel posts). Every piece is fully written, not a topic suggestion: real hooks, real captions, real hashtags, slide-by-slide carousel copy, and production notes. You choose the volume up front at 36, 60, or 90 pieces.

Say this to start

This skill has no button. You start it by saying what you want. Any of these will do it:

/build-content

> build my content system

> 30-day content sprint

> content system

> 90 pieces of content

> content calendar and pillars

When to reach for it

starting a content engine from zero
you have an offer but no publishing plan
you need 30 days of posts scheduled in advance
building a content system for a client brand
your posting is ad hoc and off-message
you want tiered content that moves cold traffic to conversion

When NOT to use it

If you actually want	Use this instead
one standalone post or caption	`social-caption-writer`
a single long-form SEO article	`blog-writer`
turning an existing asset into derivatives	`content-repurposer`
cutting clips out of existing footage	`video-repurposer`
paid ad copy	`meta-ads-copywriter`
the image or video prompts behind a piece once it is written	`creative-director`

Before you start

What you need	Why
An offer already built (`offer-output.md` in the target)	the skill blocks and offers to run /build-offer first, because content without a pinned offer drifts generic	Required
Company or agency name and landing page URL	brand voice is extracted from the live page in research phase 1	Required
Industry in `broad industry > specific niche` format	the whole research phase is scoped to this, and a vague answer produces vague pillars	Required
2 to 5 competitor names to start from	the skill expands these to 8 to 15 during research and mines their content gaps	Required
Platform selection and a priority ranking	it creates content only for the platforms you name, so an unselected platform gets zero pieces	Required
An honest answer on content capacity: Light 36, Standard 60, or Aggressive 90 pieces	it will not default to 90 without asking, and picking 90 when you can publish once a day wastes most of the output	Required
An ICA document (`ica-output.md`) or one you paste in	every hook has to call out the ICP by name, and without it the pieces read generic	Required
`epistemology.md` in the workspace	content pillars are drawn from your named distinctions and reframes; without it you get a one-line warning and a shallower plan built from owner-profile topics only	Optional
`verified-claims.md` with consent-tagged proof	any number, result, testimonial or named client that is not in the registry gets replaced with a [PROOF GAP] marker you have to fill yourself	Optional
Perplexity and Firecrawl MCP connections	these drive the competitor scrape and the live platform-trend research; without them it falls back to web search and the intel is thinner	Optional

How it runs

Prerequisite and overwrite checkResolves whether you are in flat or product mode, then checks that an offer exists. If a content plan is already there, it gives you four options: use existing, update in place, archive and replace, or cancel. It never overwrites silently.
Market intelligence sweepBlocking step before anything is written. It researches competitor topics and cadence, current industry discussion, and platform algorithm and format shifts. Content runs on the fast freshness track because platforms move in days, so this re-runs more often than other builds.
Capacity and platform interviewAsks for company name, URL, industry, ICP, offer, competitors, platforms, platform priority, and your realistic publishing capacity. It recommends a capacity level but waits for you to choose before generating anything.
Phase 1: research intelligenceFive sub-phases: landing page brand analysis, competitor expansion to 8 to 15 and deep analysis, industry leader framework extraction from 3 to 5 leaders, ICP validation against real data, and platform trend analysis. Output is Research_Intelligence_Report.md, and everything downstream cites it.
Phase 2: write the piecesThree tiers by three formats. Tier 1 cold is educational with zero product mentions, tier 2 warm builds authority with soft integration, tier 3 hot converts directly. Hooks are chosen first from seven named archetypes, before the body is written. The nine tier-and-format batches are dispatched in parallel and assembled back into consolidated files.
Quality gatesA two-tier hook gate runs a linter for placeholder text and banned openers, then a hook grader that judges whether each hook would actually stop a scroll. Tier 3 pieces and any tier 2 piece with a direct CTA also go through a conversion copy grader. A two-tier claim check catches invented numbers and invented anecdotes. Nothing is presented while any gate is failing.
Phase 3: package the deliverablesAssembles the master content library spreadsheet, the 30-day calendar with its progressive tier ratios (week 1 leans 70 percent cold, week 4 leans 40 percent hot), the platform optimisation guide, the brand voice guide, and the content creation SOP.
Six-step gate check and close-outAuto-scores the build and fixes anything below 9, saves .docx versions, uploads to Drive if the gws tool is present, updates the dashboard, asks whether you want a presentation deck, and prints your build progress. It then asks how the output landed and logs your answer.

What you get

Research_Intelligence_Report.md covering all five research sub-phases
Nine consolidated content files in content/: Tier_1_Video_Scripts.md through Tier_3_Carousel_Posts.md, with IDs T1V01 to T3C10
Master_Content_Library.xlsx: an 18-column index of every piece with tier, format, hook type, pillar, publish date and status
Content_Calendar_30_Days.xlsx scheduled to your chosen capacity
Platform_Optimization_Guide.md with per-platform length, format and hashtag rules
Brand_Voice_Guide.md and Content_Creation_SOP.md so a team member can run the system without you
content-plan.md written to the resolved target, plus an updated assets.md
Per piece: full script or caption, three hook variations each labelled with its archetype, platform-specific captions with hashtags, visual direction, and production instructions

Honest limits

Read this before you rely on it

This is the longest build in Agency OS. At 90 pieces it can outrun a single session. The skill saves finished pieces as it goes and tells you to start a new session and say 'continue content build', so plan for that rather than being surprised by it.
It writes copy, briefs and prompts. It does not produce the finished videos, images or carousel graphics. Production is a separate step, and for video scripts it will offer to hand off to the workspace's avatar video tool.
Any specific claim (a number, a result, a testimonial, a named client, a dated event, a first-person anecdote) that is not in your verified-claims registry is stripped out and replaced with a [PROOF GAP] marker. You have to supply the real proof before publishing.
Research depth is only as good as the connections you have. Without Perplexity and Firecrawl the competitor and platform-trend intel falls back to plain web search, and the market research step writes an error stub and carries on if it times out.
The output is sized to the capacity you claim, not the capacity you have. Choosing 90 pieces and publishing three a day is a real commitment or a team job. Light at 36 pieces is the honest choice for most solo operators.

Where people go wrong

The mistake	Do this instead
Picking Aggressive 90 pieces because it sounds like better value	Pick the level you will actually publish. Light is 36 pieces at three days a week and is sustainable solo. You can rerun for more later.
Running it before the offer is built	Run /build-offer first. The skill will stop and tell you this anyway, and content built without a pinned offer pulls in the wrong people.
Naming every platform to keep options open	Name only the platforms you actually post on. Content for an unused platform is wasted output, and the priority ranking is what allocates effort.
Treating hooks as something to tidy up later	The hook is chosen before the body is written and must declare one of the seven archetypes. A hook that could front any brand in any niche is treated as a defect and sent back.
Expecting the calendar to be evenly mixed	It is deliberately progressive. Week 1 is 70 percent cold content, week 4 is 40 percent conversion content. Do not reshuffle it into a flat mix.

Worth knowing

The three hook variations on each video script should span two or three different archetypes, not three rewrites of the same one. That is what gives you a real A/B test instead of three versions of the same failure.