PLGN
Browse the docs

What plgn knows about your brand

Everything the plugin writes rests on what is saved about your brand. It is kept in four different places, and putting something in the wrong one does not fail loudly: it fails months later, when a rule nobody can find is not being applied.

Reading time: 12 minutes

The four places

PlaceWhat it holds
The brand recordThe name, the languages it publishes in, the words it refuses to use, and the timezone it publishes on.
Knowledge entriesWhat the brand knows about itself: how it sounds, who it talks to, how it looks, its proof, its competitors. Seventeen kinds, in three layers.
OfferingsWhat it sells. One record per product or service, each with its benefits spelled out.
CampaignsOne thing it is saying for a while, with a start and an end, a key message and its own rules.

Topics, snippets, hashtag sets and image files are none of these. They are ordinary content with their own place in the workspace.

The three layers

Knowledge entries are grouped by how often they change, and that grouping decides what overrides what. Foundation is read first and nothing overrides it. These three words appear as headings on the Knowledge page in your dashboard, so they are worth knowing.

LayerWhat it isWhat is in it
FoundationWho the brand is. Read first, never overridden, and changing any of it needs your approval.Identity, positioning, voice, audience, visual rules, creative rules.
BusinessWhat it sells and the proof behind it. Changes as the business changes.Promotions, proof, objections, competitors, market context, search rules, platform rules, and the brand's channels (its website and social accounts).
CreativeHow it says one thing, this once.References, approved executions, example posts worth imitating.

Seventeen kinds in total, and that is the whole list. There is no "offers" kind and no "publishing" kind: those are an offering and the brand record, and that is exactly the confusion this page exists to settle.

Four things a brand has exactly one of

A brand holds exactly one identity, one positioning, one voice and one audience. Trying to add a second is refused, and the refusal names the one that already exists so it can be updated instead.

This is why running /plgn setup or /plgn brandkit twice is safe. The second run compares against what is saved and updates it rather than leaving you with two of everything.

Foundation needs your yes

Every write to a Foundation entry is shown to you first and saved only after you approve the exact text. This is not a formality. Foundation is what every future post reads, and the approval is the difference between a brand that agreed to how it sounds and one that was told.

Business and Creative entries do not need it. They are facts about the market and examples of work, and they change often enough that confirming each one would be noise.

Banned words, the one that catches everybody

If you have ever saved a banned-word list and watched posts keep using the word, this is the reason. /plgn knowledge will tell you where yours is, and /plgn setup puts it in the right place from the start.

Adding one word replaces the whole list

The list is saved whole, not appended to. The plugin reads your current list first and sends it back with the new word added. This matters if you ever edit it another way: sending one word deletes the rest.

What to put on it

  • Competitor names you never want to appear in your own posts.
  • Industry clichés your audience is tired of, the phrases that could open any post in your market.
  • Claims you cannot back up. If you cannot prove "the leading", ban it and the problem is solved once.
  • Both spellings of anything hyphenated, because the check matches text: "game-changing" and "game changing" are two entries.

The timezone

The timezone lives on the brand record, beside the languages, because it is the same decision: where this brand publishes, and therefore at what hour.

If you are not sure yours is set, /plgn knowledge reports it and /plgn setup asks for it directly. "Where do you post from?" is a question anyone can answer in one line.

What you sell

Each product or service is its own record, not a knowledge entry. That is deliberate: a writer needs to name it correctly and an art director needs to picture it, and neither works from a paragraph buried in general knowledge.

Every offering carriesWhy
A name, written the way you write itSo no post invents a different name for your own product.
Product or serviceThey carry different fields. A product has variants and pictures; a service has deliverables, a process, an outcome and how you work together.
One hero, the rest supportingOnly one offering is the hero at a time. Setting a second one demotes the first, which is deliberate and worth knowing.
Benefits, with meanings and clichés to avoidThis is the part that matters most and the part most often left empty.

A benefit with only a label is not a benefit

"Roasted weekly" is a label. What a writer needs is what it means, "never sits in a warehouse", and what not to say about it, "farm to cup". A benefit carrying only a label gives every writer a phrase to repeat, which is the opposite of what it is for.

Campaigns

A campaign is one thing you are saying for a while, with dates. It holds a key message, the words to reach for, the words to avoid, the things it is selling and the topics it covers.

  • A post written inside a campaign inherits its offerings, and its writer is given the key message and the constraints.
  • A campaign is current when it is running and today falls inside its window. Reads pick up the current ones automatically.
  • A draft campaign with no dates never becomes current, so nothing ever inherits it. If a campaign seems to be doing nothing, check its dates first.
  • Marking one done stops anything new inheriting it, and leaves its posts exactly where they are.

How the brand looks

A brand with a voice and no look publishes posts that read right and look like stock. The look is read off pictures you have already published, never invented, and it is saved as one entry with ten parts.

PartWhat it records
PaletteThe colours, with rough proportions and how backgrounds are treated
CompositionWhere the subject sits, how tight the crop is, where text can safely go
LightDirection, hard or soft, warm or cool, how shadows behave
MediumPhotograph, illustration, 3D render or collage, and the camera feel
SubjectWhat actually appears, and when there are people, who they are and what they are doing
FinishMatte or glossy, flat or gradient, texture, the colour grade
Text in the imageWhether words appear at all, where, and how heavy
NeverWhat these pictures never contain
The preambleA block put in front of every later image description
The reference pictureThe one image that best represents the set

The "never" list carries more weight than it looks. A brand whose pictures never show a face, never use pure white, or never contain a logo has an identity built on those refusals, and getting a refusal wrong is what makes a generated image feel like a different company.

Three references is the minimum. Fewer is a sample, and /plgn visuals says so rather than dressing a guess up as a direction. Where your references disagree, it shows you the groups and asks which is current instead of averaging them.

Every change is kept

Changing an entry stores the old text alongside the new one, with a short note saying what changed and who changed it. So "when did our voice change?" is a question with an answer, six months later.

This is also why entries are updated in place rather than deleted and re-added. If a second write failed, a deleted entry would be gone and the brand would be worse off than before it was touched, and deleting also throws away every stored version of it.

The Brand plan's caps

Every plan -- including the 14-day Brand trial every new workspace starts on -- sets its own limit for knowledge entries, offerings and campaigns in progress. Settings → Plan is the live source for what your plan allows: a limit changed from the HBS dashboard shows up there at once, with nothing here to go stale.

Hitting one is a refusal, not an error. You are told what is full, what did not get saved by name, and what raising the limit costs. Nothing is ever dropped in silence, and nothing is retried in a loop.