Brand Voice Dossier Builder
Extract a reusable brand and founder voice system from the organization’s own content.
The evidence limits processing to user-owned material, treats pasted content as untrusted, forbids competitor scraping, and routes canonical writes through an authorized propose event and a sole-writer registry; however, it does not fully specify minimization, retention, rollback, or withdrawal controls for sensitive emails and decks, nor confirmation before every registry proposal, so 7 points are deducted.
The procedure, NEEDS_INPUT behavior, termination conditions, citation requirements, and failure feedback are relatively clear; however, registry-events.py, referenced protocols, and connectors are not reproduced here, and there is no skill-specific test coverage or executed evidence for key paths and abnormal inputs. The static cap therefore limits this to 9.
Triggers, audience scenarios, expected outputs, non-fit boundaries, and Chinese-platform examples are clear, with user exports required for closed platforms; however, operation depends on channel state and upstream protocols, fallback behavior is incomplete when those are absent, and mainland-China reachability is not demonstrated, so 4 points are deducted.
Front matter provides name, version, Apache-2.0 licensing, compatibility, and an argument hint; the document also supplies quick starts, a contract, sources, procedure, save rules, and handoff structure. It lacks a skill-specific changelog, explicit maintenance owner and update path, FAQs, and fuller installation/dependency troubleshooting; publisher identity is also unverified, so 5 points are deducted.
The intended output is concrete: a versioned voice dossier with platform registers, banned phrases, disclosure lines, own-post exemplars, and pillar allocations, with explicit completion criteria. Static review provides no produced artifact or third-party execution evidence, so direct usability and marginal benefit over manual work remain unverified; the static cap limits this to 7.
The skill requires each trait and exemplar to cite an owned post and requires versioning, while the repository supplies some general architecture and security tests. There is no skill-specific test suite, sample output, independent reproduction record, or corroborating evidence across sources, so this receives 4 points.
- Do not write unconfirmed email, deck, or post content into persistent memory or registries; verify sensitive-data scope, retention, and proposal rollback procedures first.
- This review did not execute the skill or connectors; submission and handoff may fail if registry-events.py, referenced protocols, or host capabilities are unavailable.
- For Chinese platforms, assume user exports or screenshots are required; do not assume compliant automated retrieval.
What does this skill do, and when should you use it?
Voice Dossier Builder extracts how a brand and its founder or executives actually communicate from the user’s own posts, emails, and decks. It creates a per-platform register map, banned phrases, disclosure lines, an own-content few-shot bank, and estimated content-pillar allocations. It does not scrape competitors, invent personas, conduct audience research, or write posts. The versioned record is submitted to channel-registry for promotion and consumption by later social-content skills.
Reads the user’s own posts, emails, decks, and available first-party material; uses open own-profile or RSS sources where available and requests exports or screenshots for closed platforms; extracts diction, rhythm, emoji and hashtag habits, code-switching, recurring patterns, and taboo topics before interviewing to fill gaps; produces brand and founder platform registers, banned phrases, disclosure lines for founder/employee/advocate and AI-media contexts, a few-shot bank containing only own posts, and 3–5 content pillars whose estimated allocations total 100%; saves a working dossier and submits a registry proposal through registry-events.py.
- A brand team wants consistent voice rules for its company and founder accounts before drafting social content.
- A founder has a body of posts, emails, and decks and wants a reusable voice record for downstream content work.
- A team is adding Xiaohongshu or WeChat Official Accounts and needs a register derived from its own published material.
- A marketing team needs approved banned expressions and disclosure wording for different posting identities and AI-assisted media contexts.
- A social team needs to establish 3–5 content pillars and estimated starting allocations from its existing corpus and objective.
What are this skill's strengths and limitations?
- Uses the user’s own material and explicitly excludes competitor content and invented personas.
- Covers platform registers, banned language, disclosure lines, own-content examples, and pillar allocation.
- Requires versioned registry submission through an authorized proposal flow.
- Includes Chinese-platform contexts such as Xiaohongshu and WeChat Official Accounts.
- Requires enough first-party posts, emails, or decks; it will stop when the corpus is insufficient.
- Closed platforms cannot be read through compliant keyless retrieval and require user exports or screenshots.
- The source provides no independent test results or complete standalone installation procedure for this skill.
- It does not perform audience or persona research, maintain platform norm cards, or write social posts.
How do you install this skill?
Use the repository’s documented generic installation route: git clone https://github.com/aaron-he-zhu/aaron-marketing-skills. The source does not document a single-skill installation command for the voice-dossier-builder slug; the skill file is at social/explore/voice-dossier-builder/SKILL.md.
How do you use this skill?
Provide first-party source material in a prompt such as: Build the voice dossier for [brand]. Here are our last 30 LinkedIn posts and 10 founder tweets: [paste/export]. Another supported trigger is: Codify my founder voice from these emails and this pitch deck — I post as "I", the company account posts as "we". If the complete corpus has fewer than roughly 10 usable items, the skill stops with NEEDS_INPUT instead of inventing material.