Claude Instructions

This is the Hoover Content System (HCS), a skill-based system for content creation across UX writing, content design, content strategy, technical documentation, developer content, marketing, editorial, AI/MCP agents, and research synthesis.

What This Repo Is

HCS is a library of skills—structured packages for specific content tasks. Each skill defines exactly how to produce a type of content: what inputs you need, what workflow to follow, what format to use, and how to validate quality.

Repo Structure

/shared/              → Global content standards (apply to ALL outputs)
/governance/          → Repo rules, skill catalog, glossary, eval spec
/evals/               → Test cases for skill outputs
/{category}/{skill}/  → Individual skill packages

Skill Categories

Category	Content Types
ux-writing	Error messages, empty states, dialogs, notifications, forms, onboarding
content-design	UI copy systems, patterns, audits, naming, cognitive load
content-strategy	Voice/tone, taxonomy, messaging hierarchy, roadmaps
content-context	Feature context packs, vocabulary, state maps, user journeys
technical-documentation	How-to guides, API docs, reference docs, release notes
developer-content	CLI help, SDK docs, code comments, configuration guides
mcp-and-agents	Agent instructions, system prompts, tool descriptions, evals
marketing	Landing pages, product messaging, campaigns, app store listings
editorial-and-blog	Blog posts, case studies, thought leadership, tutorials
research-and-insights	User research synthesis, feedback analysis, audit reports

Trigger Index

Quick-match triggers to skills. See governance/skill-catalog.md for full details.

ux-writing: writing-error-messages (error, error message, failure) · writing-empty-states (empty state, no results, blank state) · writing-confirmation-dialogs (confirmation, dialog, modal, delete confirm) · writing-onboarding-steps (onboarding, welcome, getting started) · writing-settings-and-preferences (settings, preferences, options) · writing-status-and-progress (status, progress, loading) · writing-form-labels-and-helptext (form, label, placeholder, help text) · writing-permission-and-access-messages (permission, access, allow, deny) · writing-loading-and-latency-messaging (loading, wait, processing) · writing-notifications-and-toasts (notification, toast, alert, snackbar) · writing-accessible-ui-copy (accessible copy, screen reader, a11y)

content-context: creating-context-packs-for-ai (context pack, AI context) · defining-feature-vocabulary (vocabulary, terminology) · diffing-and-versioning-context-packs (context diff, version comparison) · generating-feature-content-context (feature context, content context) · generating-state-maps (state map, UI states) · mapping-content-to-user-journeys (journey mapping, user flow) · updating-feature-context-from-feedback (context update, feedback integration) · validating-context-completeness (context validation, completeness check)

content-design: auditing-ui-copy (copy audit, UI review) · designing-content-governance-flows (governance, approval flow) · designing-content-pattern-libraries (pattern library, content patterns) · designing-microcopy-systems (microcopy, UI text system) · designing-progressive-disclosure (progressive disclosure, layered content) · mapping-user-intents-to-copy (intent mapping, user goals) · naming-features-and-settings (feature naming, setting names, naming) · reducing-cognitive-load (cognitive load, simplification) · structuring-ui-content (content structure, UI organization)

content-strategy: aligning-content-with-brand (brand alignment, brand voice) · building-taxonomy-and-ia (taxonomy, IA, information architecture) · creating-content-briefs (content brief, project brief) · creating-content-principles (content principles, guiding principles) · creating-messaging-hierarchy (messaging hierarchy, message priority) · defining-voice-and-tone (voice, tone, personality) · designing-content-experiments (content testing, A/B test) · measuring-content-performance (content metrics, measurement) · planning-content-roadmaps (content roadmap, planning)

developer-content: designing-developer-onboarding (developer onboarding, quickstart) · documenting-tooling-workflows (tooling docs, workflow documentation) · writing-cli-errors-and-exit-codes (CLI error, exit code) · writing-cli-help-and-usage (CLI help, usage, --help) · writing-code-comments-and-docstrings (code comments, docstrings) · writing-command-tutorial-flows (CLI tutorial, walkthrough) · writing-configuration-guides (config guide, configuration) · writing-sdk-docs-and-samples (SDK docs, code samples)

technical-documentation: writing-how-to-guides (how-to, step-by-step, procedure) · writing-reference-docs (reference, API reference) · writing-troubleshooting-guides (troubleshooting, debug, fix) · writing-api-documentation (API docs, endpoint docs) · writing-release-notes (release notes, changelog, what's new) · maintaining-docs-as-code (docs as code, doc automation) · structuring-doc-information-architecture (doc IA, doc navigation) · documenting-breaking-changes (breaking change, migration guide) · designing-documentation-navigation (doc navigation, sidebar, TOC)

editorial-and-blog: designing-headline-variants (headline, title variants) · editing-and-style-enforcement (editing, style enforcement) · improving-readability-and-flow (readability, flow, clarity) · structuring-long-form-content (long-form structure, article structure) · writing-blog-posts (blog post, blog, article) · writing-case-studies (case study, customer story) · writing-thought-leadership (thought leadership, opinion) · writing-tutorial-articles (tutorial, how-to article)

marketing: creating-message-maps (message map, messaging matrix) · designing-conversion-copy (conversion copy, CTA) · writing-app-store-listings (app store, Play Store) · writing-email-campaigns (email campaign, newsletter) · writing-landing-pages (landing page, LP, product page) · writing-product-announcements (announcement, launch) · writing-product-messaging (product messaging, positioning, value prop) · writing-sales-enablement-content (sales enablement, battle cards) · writing-social-media-content (social media, social post)

mcp-and-agents: auditing-agent-behavior (agent audit, behavior review) · defining-guardrails-and-constraints (guardrails, constraints) · designing-agent-instructions (agent instructions, agent prompt) · designing-agent-workflows (agent workflow, orchestration) · designing-context-injection-patterns (context injection, RAG) · designing-evaluations-for-agents (agent evals, testing agents) · structuring-system-prompts (system prompt, base prompt) · writing-mcp-tool-descriptions (MCP, tool description)

research-and-insights: synthesizing-user-research (user research, research synthesis) · extracting-insights-from-feedback (feedback analysis, user feedback) · structuring-content-audit-reports (audit report, content audit findings) · writing-research-summaries (research summary, findings) · mapping-qualitative-to-quantitative (qual to quant, mixed methods) · designing-interview-guides (interview guide, user interview) · analyzing-survey-results (survey analysis, survey data) · creating-content-opportunity-maps (opportunity map, content gaps)

Skill Package Structure

Every skill folder contains:

File	What It Does	How You Use It
SKILL.md	Defines triggers, required inputs, workflow, degrees of freedom	Read first. Tells you what the skill does and what you need to execute it.
TEMPLATES.md	Provides output structures and formats	Use as your output scaffold. Follow this structure unless user overrides.
RUBRIC.md	Lists pass/fail criteria for output quality	Validate output against every criterion before delivering.
EXAMPLES.md	Shows input/output pairs of correct execution	Reference when handling edge cases or unfamiliar scenarios.

Optional files:

• CHECKLIST.md – Quick validation checklist
• reference/ – Domain-specific reference docs
• scripts/ – Validation scripts

How to Execute a Skill

Step 1: Find the Right Skill

Look up governance/skill-catalog.md. Match the user's request to a skill using triggers.

Example: User asks "write an error message for a failed upload" → triggers match writing-error-messages skill.

Step 2: Read SKILL.md

Open the skill's SKILL.md. Identify:

• Required inputs – Must have before you proceed
• Optional inputs – Have defaults if not provided
• Workflow – Steps to follow in order
• Degrees of freedom – Where you can use judgment

Step 3: Gather Missing Inputs

If the user hasn't provided required inputs, ask for them. Be specific about what you need and why.

Step 4: Apply Shared Standards

Load and apply all files from /shared:

File	Governs
shared/style.md	Capitalization, punctuation, formatting
shared/voice.md	Voice attributes, tone for context
shared/accessibility.md	Inclusive language, reading level
shared/localization.md	Translation-ready patterns
shared/legal-safety.md	Prohibited claims, disclosures

Step 5: Generate Using TEMPLATES.md

Use the structure from TEMPLATES.md as your output format. This ensures consistency.

Step 6: Validate with RUBRIC.md

Before delivering, check your output against every criterion in RUBRIC.md. All criteria must pass.

Step 7: Check EXAMPLES.md for Edge Cases

If you're unsure how to handle something, look for a similar scenario in EXAMPLES.md.

Terminology

Use canonical terms from governance/glossary.json. Never use prohibited synonyms listed there.

What You Must Not Do

• Invent rules not documented in this repo
• Skip a skill when one exists for the request
• Deliver output that fails any RUBRIC.md criterion
• Override or ignore /shared standards
• Assume file contents—always read them

Quick Start with Prompts

If the user wants a fast start, point them to PROMPTS.md. It contains ready-to-use prompts for every skill category—they just fill in the brackets and go.

When No Skill Matches

1. Tell the user no matching skill exists.
2. Apply /shared standards only.
3. Offer to help scope a new skill if the task is recurring.

Output Format

1. Lead with the content output.
2. State which skill you applied (or that none matched).
3. Flag any rubric items that may need user judgment.
4. Keep explanations brief unless asked for rationale.