Agent Instructions

This repository is an AI-native design engineering workspace.

When working in this repo:

• treat README.md as the public positioning source of truth
• use playbook/ for long-form methodology
• use skills/ for reusable workflows
• use rules/ for operating guardrails
• use commands/ for repeatable agent actions
• use prompts/ for copy-paste prompt patterns
• use doc-templates/ for reusable project artifacts
• use agents/ for role definitions
• use tools/ for scripts and helpers
• use integrations/ for setup guides and external tool workflows

Do not add private project data, client data, secrets, or personal notes.

Prefer practical, reusable, senior-level material over generic AI advice.

Available agents

• agents/agent-agile-master/AGENT.md — Agile master for solo practitioners. Routes to 25+ rituals: sprint planning, retrospective, estimation, story mapping, OKR, team health, change management. Load with PERSONA.md.

Available skills

• skills/legible-agent-output/SKILL.md — Enforce plain-language output from AI agents. Replace opaque codes (A127, ENOENT), framework jargon, raw error strings, and bare percentages with human-readable titles, status messages, and error descriptions. Cross-cutting rule for any agent emitting user-facing strings.
• skills/kano-model-strategist/SKILL.md — Feature prioritization using Kano model. Cut waste, build MDP over MVP.
• skills/socratic-dialogue/SKILL.md — Reasoning rigor and anti-sycophancy guard for high-stakes decisions.
• skills/value-proposition-canvas/SKILL.md — Run a Value Proposition Canvas workshop through Phase 1 (Customer Profile) + Phase 2 (Value Map: Pain Relievers, Gain Creators, Products & Services) + Phase 2.5 (Framework Selection). Tables as primary output. Number rules (1-2 main Jobs + 3-5 supporting, 3-5 Pains, 3-5 Gains). Explicit user acceptance gates between every step. Soft rigor (Hypothesis welcomed, not rejected). Multi-file output: produces vpc-result.md + shadow-backlog.md starter; user writes Value-proposition.md after Kano. Composes with socratic-dialogue (validation) and kano-model-strategist (next step).
• skills/typesetting-engine-skillset/ — six-skill typography/layout family (see typesetting-engine-skillset/README.md for the map): microtypography, line-length-optimizer, text-typesetting, vertical-spacing, type-scale-generator, glyph-fidelity.
• skills/ux-content-design/SKILL.md — Content design: UI microcopy, error and system messages, design system content guidelines, explainer articles, and copy audits. Writes the states that break in production and never reach the mockup. 4-phase procedure (brief → 3 variants → stress-test → ship) with four output contracts (string spec, design system content section, explainer, audit report). Marketing register out of scope by design — pastiche-tell taxonomy, golden good/bad pairs loaded before drafting, voice chart defined by constraints instead of adjectives, read-aloud gate as the first check. Bilingual PL/EN with length budgeting. Claude only. Composes with legible-agent-output (same law, machine-authored strings) and typesetting-engine-skillset (typography after the string exists).
• skills/figjam-storymap-llm/SKILL.md — Audit and parse FigJam User Story Maps (Jeff Patton methodology) into LLM-readable Markdown / JSON. Eliminates the post-workshop transcription step. Includes a Python parser (Figma REST API → Markdown / JSON), canonical FigJam template spec, verification prompt for LLM-readiness audits, and Lean UX rules (V1 has priorities + owners + AC; V2 / V3 are clean). Use after a Story Mapping workshop, before publishing a Story Map template, or when feeding a Story Map to a coding agent (Cursor, Claude Code, Copilot).
• skills/ui-roi-audit/SKILL.md — Audit web pages and task-driven interfaces against source-cited conversion, usability, and ROI criteria. Scores applicable criteria, ranks fixes by exposure, severity, effort, and expected return, and estimates conservative twelve-month revenue gain or cost avoided. Every sourced finding cites Loveday & Niehaus (2008) or Marcus (2004); bibliography in ATTRIBUTION.md. Audits only, does not draft copy. Use for "UI ROI audit", "interface audit", "usability ROI", "ROI audit", or "conversion audit".