Brand Language Codifier
Turn a brand’s published language into reusable voice, lexicon, and product-naming rules.
The skill treats pasted and fetched material as untrusted, restricts voice evidence to the brand's own material, discloses reads/writes/proposal flow, requires confirmation for memory writes, and routes canonical changes through authorized registry-events.py proposals. Repository security material also describes no credential storage, scraping controls, and fail-closed registry behavior. Deducted 4 points because scraping and memory/event writes still depend on correct host configuration, rollback and exact authorization boundaries are not fully specified in this skill, and publisher identity is unverified.
The instructions, input/output contract, NEEDS_INPUT stop condition, shipped-canon check, and termination rules are coherent, with intelligible feedback when the upstream message house is missing. Deducted 12 points because there are no skill-specific key-path tests or reproducible execution records, referenced scripts/registries/shared protocols are not verified by the supplied evidence, and abnormal-input/write-failure handling is only partial; static review caps this dimension at 10.
Triggers, scenarios, required inputs, output shape, non-fit boundaries, and upstream/downstream routing are explicit. Chinese-language descriptions are present, and the core path can use user-provided material without depending entirely on overseas services. Deducted 3 points because the skill has strong prerequisites on host memory, registries, and the upstream message house, while direct evidence for China-localized outputs and mainland network reachability is limited.
The document has layered front matter, Quick Start, Skill Contract, Data Sources, Instructions, save behavior, and Next Best Skill sections. Version, Apache-2.0 licensing, limitations, shared references, and security guidance are documented. Deducted 3 points because this skill does not clearly identify a maintenance owner, update responsibility, or changelog path, leaving governance incomplete.
The contract specifies a directly useful brand-language canon: register, tone spectrum, banned phrases, 3-6 examples from first-party material, naming rules, approved/banned term tables, and handoff data. Deducted 9 points because no actual representative output or execution evidence is supplied, and registry proposals, memory staging, and confirmation may require additional orchestration; static review caps this dimension at 7.
The skill requires source labels and Measured/User-provided/Estimated distinctions, forbids competitor-derived examples and unsupported factual claims, and links to governing protocols. Deducted 6 points because the evidence is primarily author-defined process documentation, with no skill-specific third-party corroboration, execution record, or tests covering key paths; static review caps this dimension at 5.
- The skill requires an existing durable message house; if it is absent, it should stop and route upstream rather than treating style inference as canon.
- Fetching brand material creates data-egress and network-compliance considerations; verify robots rules, authorization, privacy boundaries, and optional Firecrawl configuration.
- A canonical registry proposal is not final approval; confirm authorization, check for conflicts with the shipped canon, and validate recovery behavior if runtime writes fail.
- Trademark, comparative, and product facts in source material must not be treated as true claims; they should remain subject to claims-registry review.
What does this skill do, and when should you use it?
This skill creates a brand-level language canon covering register, tone spectrum, banned phrases, examples, and naming rules. It infers voice only from the brand’s own published material and can check an existing message house and narrative canon. It produces a voice-and-naming draft and stages canon-grade proposals for the narrative registry rather than mutating the canon directly. It fits brands with a durable message system that are preparing to scale content across channels.
Reads user-provided or keylessly scraped brand material such as homepages, documentation, blogs, decks, and social bios, plus an existing message house, narrative canon, and read-only claims ledger. Defines register, two to four tone dials, banned phrases, and three to six before/after examples based only on the brand’s own text. Establishes product, feature, and tier naming rules; creates approved-term and banned-term tables with provenance labels; marks factual or comparative claims as needing sources; and stages authorized proposals for registry review.
- A brand team has an established message house and needs a durable voice guide before scaling content production.
- A product team needs consistent rules for naming products, features, and pricing tiers.
- A marketing lead wants to derive the brand voice from published website, documentation, and blog copy.
- A multi-channel team needs platform voice adaptations to point back to one brand-level language canon.
What are this skill's strengths and limitations?
- Clearly bounded output covering voice rules and product, feature, and tier naming.
- Uses only the brand’s own published material for voice examples, reducing fabricated brand-language patterns.
- Separates drafts, registry proposals, and claims needing sources instead of adjudicating truth.
- Does not require MCP; the repository supports pasted data and keyless Tier 1 workflows.
- Requires a durable message house; without one, the workflow should stop and route upstream.
- Does not produce platform adaptations, finished copy, message hierarchy, or claim adjudication.
- Reliable voice codification depends on having sufficient first-party published material.
- Memory writes and canon proposals require an authorized workflow rather than direct file mutation.
How do you install this skill?
For the full repository in Claude Code, run /plugin marketplace add aaron-he-zhu/aaron-marketing-skills, then /plugin install aaron-marketing@aaron. On an Agent Skills-compatible host, run npx skills add aaron-he-zhu/aaron-marketing-skills; to add only this skill, use npx skills add aaron-he-zhu/aaron-marketing-skills -s brand-language-codifier.
How do you use this skill?
Provide the durable message house and the brand’s own published material. Example: Codify the brand voice for [brand] from these published samples: [paste samples]. Give me register, tone spectrum, banned phrases, and few-shots. You can also request: Build the naming tax for [product]: rules for product / feature / tier names, plus an approved-terms and banned-terms table.