Help us improve
Share bugs, ideas, or general feedback.
From aos
Build the belief / fear / JTBD map of a client's marketing decision-makers — identity, capability and value beliefs, fears, and the job each person hires marketing to do. Produces brand/BELIEF_PROFILE.md, the L0 foundation that every downstream brand file filters through.
npx claudepluginhub arcanianhq/aos-mkt-624292f82fda --plugin aosHow this skill is triggered — by the user, by Claude, or both
Slash command
/aos:aos-build-belief-profileThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
This skill's data lives in the **granted folder** — the folder Cowork was given access to, which **is** one client's folder (no per-client nesting). The granted-folder root is the working directory. Resolve zones (`client/`, `brand/`, `inbox/`, …) per `docs/data-access-router.md` and the `AOS_CONFIG.md` manifest at the granted-folder root. Never hard-code paths beyond the documented zone layout...
Mandates invoking relevant skills via tools before any response in coding sessions. Covers access, priorities, and adaptations for Claude Code, Copilot CLI, Gemini CLI.
Share bugs, ideas, or general feedback.
This skill's data lives in the granted folder — the folder Cowork was given access to, which is one client's folder (no per-client nesting). The granted-folder root is the working directory. Resolve zones (client/, brand/, inbox/, …) per docs/data-access-router.md and the AOS_CONFIG.md manifest at the granted-folder root. Never hard-code paths beyond the documented zone layout. Client identity (the client name / slug) is read from client/CLIENT_CONFIG.md and the client field of AOS_CONFIG.md — it is never a directory level. Bash + filesystem on the granted folder is the contract; the router is an optimization.
Resolve communication-language and content-language from AOS_CONFIG.md during context assembly (per docs/language-context.md) — never hard-code a language. Talk to the user in communication-language; write client-facing artifacts in content-language.
Build the Belief Profile — the beliefs driving the people who make marketing decisions in this business. Beliefs sit at L0 in the Marketing Control Framework: the source layer. Every downstream brand file — positioning, voice, offer, the GTM plan — filters through what these people believe about themselves, their capability, and what matters.
Why this is foundational. A weak-branding behaviour (won't charge a premium, associates with everyone, churns the message every quarter) is almost never a tactics problem. It traces to a belief the decision-maker holds as reality, not as a belief. This skill makes those beliefs visible so the rest of the profile can be built on truth instead of on a symptom.
The Arcanian principle (Avatar lineage):
Every business result is a creation that reveals the beliefs that created it.
We work backwards — from observable results and the decision-maker's own words to the belief that would have to be true for that result to exist.
Posture: Discovery, not pronouncement. Beliefs are surfaced as observations with questions — "this is what the evidence suggests; is this true?" — never as a verdict on someone's psychology. Present every profiled belief as a draft for the user to correct.
This skill writes exactly one file: brand/BELIEF_PROFILE.md, built to reference/output-template.md (which mirrors the aos-build-brand-system file-template). It must clear the FILLED threshold in aos-build-brand-system/reference/file-substance-criteria.md → BELIEF_PROFILE.md:
This skill operates on the granted folder — which is the client's folder. There is no client-slug argument: the granted-folder root is the working directory and client identity is read from client/CLIENT_CONFIG.md / AOS_CONFIG.md.
AOS_CONFIG.md for the zone manifest and client identity.client/CLIENT_CONFIG.md exists. If not, the folder hasn't been onboarded — suggest running aos-onboard.brand/BELIEF_PROFILE.md — even if it is a stub. The Write tool refuses to overwrite a file that hasn't been Read in-conversation. A stub returns its TODO placeholder; that is fine — the Read satisfies the harness rule.brand/7LAYER_DIAGNOSTIC.md exists and is FILLED, read its L0 and L1 sections — they often already name an identity pattern.Determine who actually makes marketing decisions in this business. Read client/CLIENT_CONFIG.md and scan the harvest for named people with decision power. For each, classify decision power: primary (owns the call), approver (signs off), influencer (shapes it). The primary decision-maker is mandatory; profile approvers and influencers when the evidence supports it.
If the harvest names no people at all — only "the company" — surface that gap and ask the user who the decision-maker is before drafting. A belief profile of a company is not a belief profile.
Scan the inbox/ zone (inbox/**/*.md, excluding inbox/_processed/) plus any <person>-belief-profile.md files elsewhere in the granted folder. Two evidence streams feed every profiled belief:
Use the result→belief mapping table and the linguistic markers in reference/belief-method.md to classify each. Tag every evidence item with its source file:line.
If the harvest yields fewer than 5 direct decision-maker quotes, the material is too thin — tell the user and ask for session logs, meeting transcripts, or a recorded conversation rather than fabricating a profile. Do not infer a person's fears from a website.
For each person, complete the five required dimensions. The method, archetype patterns, and per-dimension prompts are in reference/belief-method.md — read it before drafting:
Then identify the identity pattern (Helper/Martyr, Expert/Imposter, Visible/Invisible, Abundant/Scarce, Worthy/Unworthy — see reference/belief-method.md) where the evidence supports one, with its asserted pole, resisted pole, and core fear. State the pattern as a hypothesis, never a diagnosis.
If more than one decision-maker is profiled, map where their beliefs agree and where they conflict — and which conflicts block which marketing decisions (a contradiction between two leaders is itself a constraint). Then list the specific beliefs that block growth in the table the template provides. These are the beliefs to surface in coaching or repair work.
Present the full draft to the user with the three options:
brand/BELIEF_PROFILE.mdOnly on Accept (or post-Revise) write the file, with the frontmatter block from reference/output-template.md (scope, client, the standard provenance block — generated_by, skill_version, generated_date, aos_schema — sources_consulted, status, needs_refresh_by = generated_date + 180 days). See the Provenance section below and docs/artifact-versioning.md §1. End the file with the mandatory footer line: "What did we get wrong? What's missing?"
Every artifact this skill writes carries the standard provenance block in
its frontmatter — see docs/artifact-versioning.md §1. Stamp all four fields:
generated_by: <this skill's name> # the name: frontmatter value
skill_version: <this skill's version> # the version: frontmatter value
generated_date: <YYYY-MM-DD> # the date written
aos_schema: <schema-version> # read from AOS_CONFIG.md
Add it to whatever domain frontmatter the artifact already carries; never
hard-code skill_version or aos_schema — read them at write time. This is
what lets a granted folder be migrated when the plugin or schema changes.
file:line) — a quote or a result. No invented psychology. If a dimension has no evidence, mark it "not yet evidenced" — never fill it with a guess.scope: int-confidential — it contains sensitive observations about named individuals. Do not echo it outside the engagement.Final user-facing output:
aos-onboard (scaffolds the granted folder); aos-diagnose-7layer (its L0/L1 findings seed this profile).aos-build-brand-system consumes brand/BELIEF_PROFILE.md as a profile slot; aos-build-brand reads it when L0/L2 belief blocks resist premium pricing or visibility; aos-build-offer reads the value beliefs and fears that shape what the client will let an offer claim; voice work reads it for the why-they-talk-this-way.What did we get wrong? What's missing?