/discovery

You are running the FWS Client Discovery master workflow. This orchestrates all 9 steps sequentially, Phase A builds the website (research + page content), Phase B plans and produces the content strategy (content plan + blog posts), passing context between steps via a shared discovery-context.md file.

Research Tool Cascade

For ALL online research in every step, use tools in this order. Fall through only on failure or empty result.

1. Apify (Docker MCP), primary
   • mcp__MCP_DOCKER__apify--website-content-crawler, crawl client and competitor sites
   • mcp__MCP_DOCKER__apify--google-search-scraper, SERP, PAA, keyword discovery
   • mcp__MCP_DOCKER__apify--rag-web-browser, open-ended research queries
2. Perplexity (Docker MCP), fallback
   • mcp__MCP_DOCKER__perplexity_research, deep topic research with citations
   • mcp__MCP_DOCKER__perplexity_ask, quick lookups
3. Built-in WebSearch / WebFetch, final fallback

Log which tool produced each finding in discovery-context.md under ## Research Sources.

Output Rules (apply to every file you write)

• Markdown only. NO Word, NO Google Docs, NO XLSX, NO Pencil .pen files. Every deliverable is a plain .md file.
• No em dashes. Use a comma instead. Never write — in content.
• Internal links between pages. Every generated page and report must link to related pages and reports in the same project using relative markdown links (e.g. [Buyer Personas](./04-Buyer-Personas.md)). Run a link pass at the end of Step 7 and Step 9.

Initial Setup

1. Parse Inputs

• Client name: $1 (required, used for folder naming)
• Client URL: $2 (optional, pass none if no existing website)
• Meeting transcript: $3 (optional, file path to transcript)

1a. Preflight, ask the user

Before running any research, ask these two questions in order. Wait for each answer.

Q1, client website

"Do we have the current client's website URL? If yes, paste it now so I can crawl it for more context. If none, reply skip."

If provided and different from $2, use the newer URL and note it in discovery-context.md.

Q2, brand logo

"Do we have the client's logo? If yes, upload the file now (drag into chat or paste a path), I'll use it to extract colors and inform Step 5 UX/UI Research. If none, reply skip."

If logo provided:

• Save path to discovery-context.md under ## Brand Assets
• Step 5 reads the image, extracts dominant colors, shape language, typographic cues
• Feed into UX/UI Research report

If skip, proceed without logo analysis. Note "no logo provided" in Step 5 output.

2. Create Output Folder

Create the output directory structure:

[Client-Name]/01-Discovery/

3. Initialize Discovery Context

Copy the context template to the output folder:

• Read @${CLAUDE_PLUGIN_ROOT}/templates/discovery-context-template.md
• Write it as discovery-context.md in the output folder
• Fill in: client name, URL, industry (ask user if not clear from transcript)

4. Process Meeting Transcript

If a meeting transcript file is provided:

• Read the file
• Extract key signals: industry, services, target audience, competitors mentioned, pain points, questions
• Write findings to 01-Client-Brief.md
• Update Step 0 section of discovery-context.md

If no transcript is provided:

• Ask the user to paste the meeting notes or describe the client briefly
• Process the response the same way

---

Execute 9-Step Workflow

Run each step sequentially. After each step, re-read discovery-context.md to confirm findings landed, then proceed. Each step reads its own SKILL for full instructions (lazy-load: do NOT load all SKILLs up front).

Phase A, Research + Page Content (Steps 1-7)

#	Step	SKILL (read before starting)	Output file	Template
1	Sitemap Analysis	skills/sitemap-analysis/SKILL.md	02-Sitemap-Report.md	templates/sitemap-report.md
2	Competitor Research	skills/competitor-research/SKILL.md	03-Competitor-Report.md	templates/competitor-report.md
3	Buyer Personas (APEX)	skills/buyer-persona-research/SKILL.md	04-Buyer-Personas.md	templates/persona-card.md
4	Keyword Research	skills/keyword-research/SKILL.md	05-Keyword-Research.md	templates/keyword-report.md
5	UX/UI Research	skills/ux-ui-research/SKILL.md + CLAUDE_DESIGN.md	06-UX-UI-Research.md	templates/ux-research-report.md
6	FAQ Research	skills/faq-research/SKILL.md	07-FAQ-Research.md	templates/faq-report.md
7	Page Copywriting	skills/page-copywriter/SKILL.md	pages/*.md	(schemas in SKILL refs)

For each step:

1. Read the SKILL end-to-end, follow its workflow. SKILL loads its own references (progressive disclosure).
2. Run the research-tool cascade defined in CLAUDE.md (Apify → Perplexity → WebSearch).
3. Write the output file, update discovery-context.md, emit a one-line checkpoint to the user with key counts (e.g., "Step 1, [N] pages, [M] gaps").

Step 5 special: if logo was uploaded in preflight, run logo-analysis sub-step in the SKILL. Output MUST include a ## Claude Design Brief section per CLAUDE_DESIGN.md.

Step 7 special: end with an internal-link audit pass (3+ relative links per page).

Phase B, Content Strategy + Blog (Steps 8-9)

#	Step	SKILL	Output file	Template
8	Content Planning	skills/content-planning/SKILL.md	08-Content-Plan.md	templates/content-plan.md
9	Blog Copywriting	skills/blog-copywriter/SKILL.md	pages/blog/*.md	(schemas in SKILL refs)

Step 9 special: every post links to parent pillar + 1 service page + related posts. End with a Related Posts cross-link pass.

---

Final Assembly

Generate Master Discovery Report

• Read @${CLAUDE_PLUGIN_ROOT}/templates/discovery-report.md
• Compile executive summary from all 9 steps
• Write 00-Discovery-Report.md with:
  • Executive summary (1 paragraph)
  • Key findings per research section (Steps 1-6, 1-2 sentences each)
  • Page copywriting summary (Step 7: page count, word count)
  • Content strategy summary (Steps 8-9: content plan, blog count, topic clusters)
  • Top 5 priorities across all dimensions
  • Recommended next steps
  • Deliverables checklist

Generate llms.txt + llms-full.txt (GEO, LLM crawlers)

Follow @${CLAUDE_PLUGIN_ROOT}/skills/llms-txt-generator/SKILL.md.

• Write pages/llms.txt, the short manifest: site name, one-line description, priority links (homepage, services, about, contact, FAQ, key pillars, key blog posts).
• Write pages/llms-full.txt, full plain-text content dump of every generated page, concatenated with clear # [Page Title] separators. This is the content LLM crawlers cite.
• Both files are markdown-compatible plain text, UTF-8, no front matter.

Completion Summary

Present the user with:

1. List of all deliverables created (with file paths):
   • Master discovery report (00)
   • 7 research reports (01 through 07)
   • Content plan (08)
   • Page content files (homepage, services, about, etc.)
   • Blog post files
2. Content totals: total pages generated, total blog posts, total word count
3. Top 5 overall priorities from the discovery
4. Recommended immediate next steps
5. Note that all files are in the [Client-Name]/01-Discovery/ folder

---

Error Recovery

• If any step fails, log the error and continue with the next step
• Mark failed steps in the discovery report
• At the end, list any steps that need to be re-run
• Each step can be re-run individually using its dedicated command (e.g., /sitemap, /competitors)