Agentic Skills

A set of composable governance skills for AI coding agents working on web products and digital experiences — from initial task scoping through product design, creative direction, visual design, UX writing, promotional copywriting, implementation, and independent verification.

Each skill is a self-contained SKILL.md file under .agents/skills/<name>/, following the structure defined by skill-authoring-standard: YAML frontmatter with explicit Trigger immediately for / DO NOT trigger for conditions, a defined authority boundary, an ordered procedure, evidence requirements, validation gates, and a conflict-resolution hierarchy. Agents configured to discover SKILL.md files under .agents/skills/ can consume these directly without a separate build step.

Why this exists

Disciplined execution of individual skills (don't invent product behaviour in visual design, don't write final copy in product design, don't fabricate claims in promotional copy, don't skip verification evidence) is necessary but not sufficient — it can still produce a competent, correctly-behaved, but completely generic result: another SaaS dashboard, another card grid, another shadcn-default composition. This framework adds a layer specifically to prevent that: creative-direction owns the expressive thesis and cross-disciplinary coherence that makes a product or site feel authored and specific rather than an interchangeable assembly of familiar patterns, without giving up rigor, evidence, or authority boundaries anywhere else.

It's also deliberately not SaaS-first. Ordinary application/product work and public, expressive, editorial, or campaign web work are treated as two legitimate, coexisting routes, not one default with the other bolted on.

Installed skills

Skill	Owns
skill-authoring-standard	The meta-standard governing how every other skill is structured and validated.
task-framing	Scoping incoming requests into a bounded task frame and routing to the correct downstream discipline — including which of the two routes below applies.
product-design	Product behaviour, user workflows, information architecture, states, and product-level acceptance criteria.
creative-direction	The approved expressive thesis, anti-references, signature devices, and cross-disciplinary coherence rules that make an experience specific and authored.
visual-design	Visual hierarchy, composition, typography, colour, spacing, component appearance, and responsive/state treatment.
ux-writing	Exact interface language — labels, errors, validation, confirmations, empty states — within approved behaviour and voice.
copywriting	Promotional, persuasive, campaign, editorial, and public-facing brand-voice language — headlines, hero statements, narrative, and CTAs — within approved creative direction.
software-development	Implementing approved, bounded software changes with scope discipline and implementation self-validation evidence; supplies the governing implementation baseline frontend-development operates under.
frontend-development	Specialist frontend implementation — markup, styles, components, responsive/client-side behaviour, server-rendered frontend concerns, accessibility mechanics, performance, and approved motion — activating jointly with software-development.
testing-and-verification	Independent, evidence-backed verification of implementation claims, defects, and acceptance criteria.

How work routes through the skills

task-framing picks the route based on the task's dominant purpose — it does not default to a product-design-led pipeline merely because the deliverable is a website.

Interactive product / application
task-framing → product-design → creative-direction (when activated) → visual-design + ux-writing
    → software-development baseline + frontend-development specialist execution → testing-and-verification
      (when independently required)

Expressive / editorial / campaign / portfolio / public website
task-framing
    ├─ creative-direction, for the overall expressive concept
    └─ product-design, only where workflows, IA, forms, transactions, permissions, or interaction-state
       logic require definition
           ↓ (both branches feed forward)
    applicable visual-design / ux-writing / copywriting outputs (not every surface requires all three —
    a purely editorial page with no interface controls may need no ux-writing at all)
    → software-development baseline + frontend-development specialist execution → testing-and-verification
      (when independently required)

Mixed website (e.g. a campaign page with one signup form)
task-framing
    ├─ creative-direction, for the overall expressive concept
    └─ product-design, narrowly, only for the interactive element's behaviour
           ↓ (both branches feed forward)
    visual-design + ux-writing/copywriting consume the applicable approved outputs
    → software-development baseline + frontend-development specialist execution → testing-and-verification
      (when independently required)

frontend-development activates jointly with software-development for any production change to a frontend implementation surface, regardless of visual ambition — software-development supplies the governing baseline (scope, classification, evidence honesty); frontend-development supplies frontend-specific execution and self-validation within it. A bounded frontend task may also proceed directly from the task frame's requirements and established conventions when no upstream product-design, creative-direction, or visual-design package is necessary. Independent verification by testing-and-verification is required when the task frame, risk, or acceptance criteria call for it; otherwise self-validated work returns to the current workflow.

creative-direction grounds itself in an approved product-design specification when one exists, or directly in the task's audience, communication objective, content requirements, and user-approved references when it doesn't (public/expressive work with no product-design specification to consume).

Status

• Ten foundational skills installed; all ten currently pass skill-authoring-standard's structural validation (required sections, frontmatter, anchors) — no known structural compliance gaps remain.
• Structural validation is not the same as real-world validation: no further architectural changes are planned unless real use surfaces a concrete defect, and real-world validation across the project categories in ROADMAP.md is the active phase, not yet complete.
• Motion implementation remains inside frontend-development; whether it warrants an independent motion-development skill is evaluated through real implementation evidence gathered during that active phase, not resolved in advance.

Using these skills

Copy .agents/skills/ into a project's root. Each SKILL.md's frontmatter (Trigger immediately for / DO NOT trigger for) tells a skill-aware agent when to load it. skill-authoring-standard governs how to add, modify, or audit any skill in this directory — read it before changing the structure of an existing skill or adding a new one.

This repository's root AGENTS.md governs maintenance of the Agentic Skills framework itself — it is not automatically the operating contract for a project that consumes these skills. A host project should supply its own workspace rules appropriate to that project; the reusable package remains .agents/skills/.