gsp-project-research

Phase 2 of the GSP project diamond. Deep research phase that investigates UX patterns, competitor experiences, technical approaches, accessibility strategies, and content patterns specific to what this project is building.

This is NOT brand-level discovery (that's /gsp-brand-research). This is project-level research — focused on the product type, user flows, and implementation challenges.

Works with the dual-diamond architecture: reads brand context from .design/branding/{brand}/ via brand.ref, reads/writes project assets in .design/projects/{project}/.

Deep research into UX patterns, competitor experiences, and technical approaches for this project.

Input: Brief scope + brand system + project BRIEF.md
Output: {project}/research/ (6 research chunks + INDEX.md)
Agent: gsp-project-researcher

@${CLAUDE_SKILL_DIR}/../../templates/phases/research.md </execution_context> ## Step 0: Resolve project and brand

If .design/projects/ does not exist: output "No GSP project found. Run /gsp-start to begin." and stop.

Resolve project from .design/projects/ (one → use it, multiple → ask). Set PROJECT_PATH.

Read {PROJECT_PATH}/brand.ref → set BRAND_PATH.

Step 1: Load context

Brief (chunk-first)

Read {PROJECT_PATH}/brief/INDEX.md. If it exists, load scope.md and target-adaptations.md.

If brief doesn't exist, tell the user to run /gsp-project-brief first.

Brand system (selective)

Read {BRAND_PATH}/patterns/INDEX.md. If it exists, load foundation chunks (to understand the design system constraints).

Brand discovery (selective)

Read {BRAND_PATH}/discover/INDEX.md. If it exists, load competitive-audit.md and trend-analysis.md (to avoid duplicating brand-level research).

Brand strategy voice (for content-strategy chunk)

Read {BRAND_PATH}/strategy/voice-and-tone.md and {BRAND_PATH}/strategy/messaging.md (if exists). The content-strategy chunk derives microcopy/tone direction from brand strategy — never invent voice in isolation.

Expertise references (load on demand for chunks below)

The researcher's chunks should consult the canonical owners rather than re-deriving rules:

• accessibility-patterns.md — read ${CLAUDE_SKILL_DIR}/../gsp-accessibility/SKILL.md for WCAG criteria framing; do not duplicate WCAG specifics
• technical-research.md — read ${CLAUDE_SKILL_DIR}/../gsp-style/styles/INDEX.yml to align stack-specific token wiring with available presets

Custom references

If {PROJECT_PATH}/references/INDEX.md exists, load relevant references (competitor screenshots, brand guidelines, design specs). Pass to the researcher agent for context.

Project context

Read:

• {PROJECT_PATH}/BRIEF.md — what we're building, platforms, audience
• {PROJECT_PATH}/config.json — get implementation_target, platform, tech_stack

Step 1.5: Scope check

If design_scope is tokens:

1. Update {PROJECT_PATH}/STATE.md — set Phase 2 (Research) status to skipped
2. Display: "Research phase skipped — design scope is tokens."
3. Route: "Run /gsp-project-build."
4. Stop here.

Step 1.75: Pre-fetch references (background)

If competitor URLs or reference sites are mentioned in BRIEF.md or {PROJECT_PATH}/references/, use WebFetch with run_in_background: true to pre-fetch them. This warms content for the researcher agent.

Caps: max 5 competitor URLs per project, max 3 doc URLs per technical area. Pre-fetched content is inlined into the agent prompt — the agent does not run open-ended WebSearch during execution.

Step 2: Spawn project researcher

Load agent methodology

Read ${CLAUDE_SKILL_DIR}/methodology/gsp-project-researcher.md. Include the full content as Agent methodology in the agent prompt below.

Spawn the gsp-project-researcher agent. Inline all content — the agent should not need to read any input files.

Pass in the agent prompt:

• Content of brief scope chunks: scope.md, target-adaptations.md (loaded in Step 1)
• Content of brand patterns foundation chunks (loaded in Step 1)
• Content of brand discovery chunks: competitive-audit.md, trend-analysis.md (loaded in Step 1)
• Content of brand strategy: voice-and-tone.md, messaging.md (loaded in Step 1) — drives content-strategy chunk; never invent voice
• Content of custom references (loaded in Step 1)
• Content of BRIEF.md (loaded in Step 1)
• Any pre-fetched reference content (from Step 1.75)
• Agent methodology (loaded above)
• Research output template (from execution_context)
• implementation_target, platform, tech_stack
• Output path: {PROJECT_PATH}/research/

The agent researches using WebSearch and writes chunks directly:

• research/ux-patterns.md
• research/competitor-ux.md
• research/technical-research.md
• research/accessibility-patterns.md
• research/content-strategy.md
• research/reference-specs.md
• research/recommendations.md
• research/INDEX.md

Step 2.5: Write exports

Update {PROJECT_PATH}/exports/INDEX.md:

• If INDEX.md doesn't exist, copy it from templates/exports-index.md
• Replace everything between <!-- BEGIN:research --> and <!-- END:research --> with populated table:

<!-- BEGIN:research -->
| Section | File |
|---------|------|
| UX Patterns | [ux-patterns.md](../research/ux-patterns.md) |
| Competitor UX | [competitor-ux.md](../research/competitor-ux.md) |
| Technical Research | [technical-research.md](../research/technical-research.md) |
| Accessibility Patterns | [accessibility-patterns.md](../research/accessibility-patterns.md) |
| Content Strategy | [content-strategy.md](../research/content-strategy.md) |
| Reference Specs | [reference-specs.md](../research/reference-specs.md) |
| Recommendations | [recommendations.md](../research/recommendations.md) |
<!-- END:research -->

Step 3: Update state

Update {PROJECT_PATH}/STATE.md:

• Set Phase 2 (Research) status to complete
• Record completion date

Step 4: Phase transition output

Invoke /gsp-phase-transition with phase research and output directory {PROJECT_PATH}/research/.