docs(claude-config): business evaluation prompts D/E/F + session close 2026-09-09

- config/prompts: D API/RapidAPI, E KDP, F Etsy — viability + marketing plan agents (no social media, no cold outreach, opt-in only, 0 cap) - context.md: 00:10 current-value block; credits corrected to $38.55 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01X9iCzmxK2zbb8H1Ld3f8AN
2026-09-09 00:04:29 -05:00
parent 3cb1796338
commit dc616f83fa
5 changed files with 153 additions and 6 deletions
@@ -1,10 +1,9 @@
 # Claude & Claude Code — Research & Config Home — Context

-> **NEXT SESSION PLAN — CURRENT VALUE (rewritten 2026-09-08 23:45, Fable 5.1, supersedes the 23:05 block, now history).** Fable credits **$38.55 used** per user 23:49 (my ~$16 estimate was wrong; expire 2026-09-19). Research track for tonight is COMPLETE; everything below is decided and recorded.
-> 1. **Harness track CLOSED** (`decisions/DECISIONS.md` 22:50, 23:20, 23:32 entries). Baseline wins: keep Claude Code + our layer; LifeOS/jcode/all OSS executors rejected (Anthropic legal page confines Pro OAuth to the unmodified Claude Code binary, enforced since 2026-04-04); no trial; voice = locked bespoke plan; **user: stay on Claude Code until hybrid brain + trained local brain + affordable claude -p/API, and only move if skills/processes port.** Sonnet 5 = native 1M on Pro (verified) → row 245 window-aware hook, LOW priority. Steal rows 242–244 (Doctor check, keyword→skill injector, MEMORY.md defrag). Open-source/open-core: rows 241 (parked), 246 (design the OSS-first evaluation pipeline), business_ideas 28 (parked).
-> 2. **HyperAgent ABANDONED** (23:42 entry; business row 41 closed). **NEW Horizon-A priority:** business_projects 42 DONE 23:55 (verdict NO — GitHub cold email retired, `research/github-aup-lead-pipeline.md`) → the **marketing-system overhaul row** (all digital businesses; user is not a salesman) — do both in the API Idea conversation; write the channel-research prompt into `config/prompts/` + Obsidian `Resources/Prompts/`.
-> 3. **agent-wrap skill + SubagentStop hook built 23:05** (`/opt/appdata/docker/.claude/skills/agent-wrap`, `hooks/subagent-wrap-reminder.py`, settings.json backup .bak.20260908; hook fires from next session). Three runs logged in `/opt/appdata/docker/.claude/logs/agent_runs.jsonl`; anti-pattern added (corrections go INTO the report as a banner). **/opt/appdata/docker has UNCOMMITTED hook/skill edits** (automation-detector.py, session-start.sh, subagent-wrap-reminder.py, skills/agent-wrap/) — commit at that repo's checklist.
-> 4. **Reports:** `research/ai-harness-layer-research.md` (banner: "/recall broken" claim was FALSE), `research/jcode-vs-claude-code.md`, `research/hyperagent-marketing-fit.md`; INDEX rows present. Skills track (rows 223–237) unchanged; Opus 4.8 executes GAMEPLAN Step 0 + skills 223→224→210→225 later.
+> **NEXT SESSION PLAN — CURRENT VALUE (rewritten 2026-09-09 00:10, Fable 5.1, supersedes the 23:45 block, now history).** Fable credits **$38.55 used** (user-reported; my estimates ran ~half of actual — check the real balance first). Credits expire 2026-09-19.
+> 1. **NEXT: spend remaining credits evaluating the three launch businesses + their marketing plans.** Prompts WRITTEN 00:05: `config/prompts/D_*` (API/RapidAPI, business_projects 44), `E_*` (KDP J.M. Hartley, row 45), `F_*` (Etsy ConfettiPrintCo, row 46); mirrored to Obsidian `Resources/Prompts/`. Spawn **D → E → F one at a time**, run the `agent-wrap` skill after each (SubagentStop hook now reminds), plain-terms follow-up grill-me per report, DECISIONS.md entry per business. Outputs `research/{api,kdp,etsy}-business-evaluation.md` + INDEX rows. All three feed row 43 (marketing overhaul). Owner constraints baked into the prompts: no social media, no cold outreach, no logged-in automation, $20 cap, opt-in only.
+> 2. **Decided 2026-09-08 (all in DECISIONS.md):** harness = stay on Claude Code (LifeOS/jcode/OSS executors rejected; Pro OAuth confined to the Claude Code binary; user: revisit only after hybrid brain + local brain + affordable claude -p/API, and only if skills port); HyperAgent abandoned; **GitHub-sourced cold email RETIRED** (AUP §7/ToS §H, `research/github-aup-lead-pipeline.md`; workflow 4fuzFJAclba2Syy7 stays paused). Initial API-marketing ideas are in row 43 notes.
+> 3. **Built 2026-09-08:** `agent-wrap` completion-checklist skill + SubagentStop hook (`/opt/appdata/docker/.claude/`, committed there 00:10); runs in `logs/agent_runs.jsonl`. Rows 240–246 (agent-wrap, open-source parked, 3 steal ideas, window-aware hook low, open-core evaluation pipeline); business_ideas 28 parked. Skills track 223–237 unchanged; Opus 4.8 later executes GAMEPLAN Step 0 + skills 223→224→210→225.

 ## What this project does
 The dedicated home for everything about **how we use Claude and Claude Code to do work** —
@@ -0,0 +1,49 @@
+# Prompt D — API business (RapidAPI): viability evaluation + marketing plan
+
+Tracks: business_projects 44. Spawn from `/opt/appdata/docker/Business/API Idea` or claude-config. Research-only.
+
+---
+
+You are a business analyst evaluating a one-person API business and writing its go-to-market plan. You will read the web and local files and write exactly one Markdown file.
+
+## Task
+(1) Judge whether selling APIs on RapidAPI can produce the first paying subscribers for this owner, (2) pick the category and the next API to build using demand evidence, (3) write a marketing plan that needs no social media and no cold outreach.
+
+## Context / pre-fetched data
+- Current state: one API listed, Dividend Tracker (dividend-tracker-api), FastAPI on Coolify. Prior marketing = GitHub-sourced cold email: 4 free-tier subscribers total, 0 paid; that channel is retired for policy reasons. Owner's hypothesis: the first API was too niche; RapidAPI's ranking becomes passive once the first paid subscribers exist.
+- Owner's own initial ideas (2026-09-08, business_projects 43 notes): pick by demand via RapidAPI category/popularity data; make the listing the product (promise, free tier, runnable examples, latency, changelog); doc pages + example repo for search traffic; product-led free→paid with an opt-in 80%-quota email; consider additional marketplaces.
+- Read first: `/opt/appdata/docker/Business/API Idea/.claude/context.md` and `research/github-aup-lead-pipeline.md`.
+
+## Constraints (apply to every recommendation)
+- Owner is one person, not a salesman, with NO social media presence and no wish to build one. Any channel that requires a daily posting persona is out; evergreen, automatable, or pay-per-result channels only.
+- Money: no Anthropic API key (Claude Pro only; `claude -p` on SDK credits, $20/mo hard cap); self-hosted tooling (N8N orchestration, Postgres, local Ollama llama3.1:8b + nomic-embed-text on an 8 GB GPU). Ad budgets must be stated as explicit numbers the owner can approve.
+- Platform terms: no automation of logged-in accounts (Amazon, Etsy, LinkedIn, GitHub) and no cold email from scraped data (GitHub AUP §7 already retired the old lead pipeline, 2026-09-08). Opt-in contact only.
+- Every recommendation must say: what it costs, how long until first signal, how it is measured, and which parts N8N + a local model can automate. Every automation must name where its decisions get logged as training data.
+- Research-only: no installs, no config changes, no DB writes, no credentials, no commits. Every claim has a URL and date; write UNVERIFIED rather than guess.
+
+## Step-by-step
+1. Read the two local files (one turn).
+2. RapidAPI demand: how the marketplace ranks/promotes listings (official docs/blog only), what public data exists per listing and category (subscriber counts, pricing, popularity sort), and whether an API-side scrape of category pages is permitted by RapidAPI's terms (cite). Two turns.
+3. Category scoring: propose a scoring method (paid-subscriber density, price points, incumbent weaknesses: uptime, docs, latency, last update) and apply it to at least 8 categories from public data; shortlist 2 API ideas the owner could build in ≤2 weeks with FastAPI. Three turns.
+4. Marketing plan: the listing checklist; 3 evergreen search-driven assets (doc pages, example repo, Q&A answers); free→paid conversion design; other marketplaces worth a listing (cite terms/fees); a first-30-days calendar with weekly hours. Two turns.
+5. Viability: cost, time-to-first-paid-subscriber estimate with evidence from comparable sellers, and the kill criteria (what result after 60 days means stop).
+6. Write `research/api-business-evaluation.md` with sections: Verdict; Category scoring table; Two API candidates (what, why, build size, pricing); Marketing plan (listing checklist, search assets, conversion, marketplaces, 30-day calendar); What N8N/local model automate + training-log table; Kill criteria; Grill-me questions (≤6); Sources. Append one row to `research/INDEX.md`.
+
+## MANDATORY WRAP-UP (required regardless of success or failure)
+
+Before stopping for ANY reason — task complete, error, or approaching turn limit — output this JSON as your final message. Do not stop without it.
+
+{
+  "status": "succeeded|partially_succeeded|failed",
+  "actions_taken": ["action 1 — outcome", "action 2 — outcome"],
+  "actions_failed": ["action — reason it failed"],
+  "notes": "anything relevant for the next session",
+  "viability_verdict": "go|go_with_changes|no_go + one sentence",
+  "first_customer_path": "<one sentence>",
+  "monthly_budget_needed_usd": "<number or 0>"
+}
+
+If you hit --max-turns before finishing, set status="partially_succeeded" and list what remains in actions_failed.
+
+--max-turns 15
+If you issue the same tool call or command twice with identical arguments, STOP immediately and output the mandatory wrap-up with status=partially_succeeded.
@@ -0,0 +1,48 @@
+# Prompt E — Amazon KDP (J.M. Hartley, AI-assisted Q&A books): viability evaluation + marketing plan
+
+Tracks: business_projects 45. Spawn from claude-config. Research-only.
+
+---
+
+You are a business analyst evaluating a one-person Amazon KDP publishing venture and writing its marketing plan. You will read the web and local files and write exactly one Markdown file.
+
+## Task
+(1) Judge whether niche question-and-answer style books (the "100k Whys" format), produced with Claude + Canva and published print-on-demand under the pen name J.M. Hartley, can reach first sales for this owner; (2) define how to pick niches by demand; (3) write a marketing plan with no social media and no cold outreach.
+
+## Context / pre-fetched data
+- Business idea record: business_ideas 14 "AI-Generated Q&A Books on Amazon KDP" (status researching); launch row business_projects 40 "first 3 titles" (in_progress); target start 2026-09-14. Zero inventory; Amazon prints and ships; royalty up to 60% (verify current rates).
+- Owner concern: every digital business needs customers and the owner is not a salesman.
+
+## Constraints (apply to every recommendation)
+- Owner is one person, not a salesman, with NO social media presence and no wish to build one. Any channel that requires a daily posting persona is out; evergreen, automatable, or pay-per-result channels only.
+- Money: no Anthropic API key (Claude Pro only; `claude -p` on SDK credits, $20/mo hard cap); self-hosted tooling (N8N orchestration, Postgres, local Ollama llama3.1:8b + nomic-embed-text on an 8 GB GPU). Ad budgets must be stated as explicit numbers the owner can approve.
+- Platform terms: no automation of logged-in accounts (Amazon, Etsy, LinkedIn, GitHub) and no cold email from scraped data (GitHub AUP §7 already retired the old lead pipeline, 2026-09-08). Opt-in contact only.
+- Every recommendation must say: what it costs, how long until first signal, how it is measured, and which parts N8N + a local model can automate. Every automation must name where its decisions get logged as training data.
+- Research-only: no installs, no config changes, no DB writes, no credentials, no commits. Every claim has a URL and date; write UNVERIFIED rather than guess.
+- KDP-specific: Amazon's current AI-generated-content disclosure rules and content guidelines must be quoted and applied; no review manipulation; Amazon Ads is allowed and is the main paid lever — model it.
+
+## Step-by-step
+1. KDP rules 2026: AI content disclosure, low-content/medium-content book policies, royalty rates and print costs, category/keyword limits. One or two turns, official pages only.
+2. Niche selection method: how to read Amazon demand without paid tools (BSR thresholds, category best-seller lists, search-suggest, competitor counts) and apply it to at least 8 candidate niches for Q&A books; shortlist 3. Three turns.
+3. Marketing plan: on-Amazon levers (title/subtitle/keywords/categories/A+ content/Look Inside), Amazon Ads starter structure with a stated monthly budget and expected ACOS, launch pricing, series strategy, and 2 off-Amazon evergreen assets that need no persona (author site page, email opt-in inside the book). Two turns.
+4. Viability: unit economics per title (print cost, price, royalty, ad cost per sale), time-to-first-sale evidence from comparable publishers, and kill criteria after 3 titles / 90 days.
+5. Write `research/kdp-business-evaluation.md` with sections: Verdict; Rules that bind us (quotes); Niche scoring table + 3 shortlisted niches; Marketing plan (on-Amazon, Ads budget model, off-Amazon assets, 30-day calendar); What N8N/local model automate (niche scans, keyword tracking, sales logging) + training-log table; Kill criteria; Grill-me questions (≤6); Sources. Append one row to `research/INDEX.md`.
+
+## MANDATORY WRAP-UP (required regardless of success or failure)
+
+Before stopping for ANY reason — task complete, error, or approaching turn limit — output this JSON as your final message. Do not stop without it.
+
+{
+  "status": "succeeded|partially_succeeded|failed",
+  "actions_taken": ["action 1 — outcome", "action 2 — outcome"],
+  "actions_failed": ["action — reason it failed"],
+  "notes": "anything relevant for the next session",
+  "viability_verdict": "go|go_with_changes|no_go + one sentence",
+  "first_customer_path": "<one sentence>",
+  "monthly_budget_needed_usd": "<number or 0>"
+}
+
+If you hit --max-turns before finishing, set status="partially_succeeded" and list what remains in actions_failed.
+
+--max-turns 15
+If you issue the same tool call or command twice with identical arguments, STOP immediately and output the mandatory wrap-up with status=partially_succeeded.
@@ -0,0 +1,48 @@
+# Prompt F — Etsy (ConfettiPrintCo, AI-designed POD + printables): viability evaluation + marketing plan
+
+Tracks: business_projects 46. Spawn from claude-config. Research-only.
+
+---
+
+You are a business analyst evaluating a one-person Etsy shop and writing its marketing plan. You will read the web and local files and write exactly one Markdown file.
+
+## Task
+(1) Judge whether AI-generated designs sold as print-on-demand goods (via Printify) and/or printable digital products (journal supplies, junk-journal kits) under the shop ConfettiPrintCo can reach first sales for this owner; (2) decide POD vs digital vs both; (3) write a marketing plan with no social media and no cold outreach.
+
+## Context / pre-fetched data
+- Business idea record: business_ideas 17 "AI Print-on-Demand + Digital Products on Etsy" (researching); launch row business_projects 39 "ConfettiPrintCo Etsy launch" (in_progress); target start 2026-09-14. Design tools: Google Flow / Nano Banana + Canva; fulfilment Printify.
+- Owner concern: not a salesman, no social media; wants Etsy's own search to do the selling.
+
+## Constraints (apply to every recommendation)
+- Owner is one person, not a salesman, with NO social media presence and no wish to build one. Any channel that requires a daily posting persona is out; evergreen, automatable, or pay-per-result channels only.
+- Money: no Anthropic API key (Claude Pro only; `claude -p` on SDK credits, $20/mo hard cap); self-hosted tooling (N8N orchestration, Postgres, local Ollama llama3.1:8b + nomic-embed-text on an 8 GB GPU). Ad budgets must be stated as explicit numbers the owner can approve.
+- Platform terms: no automation of logged-in accounts (Amazon, Etsy, LinkedIn, GitHub) and no cold email from scraped data (GitHub AUP §7 already retired the old lead pipeline, 2026-09-08). Opt-in contact only.
+- Every recommendation must say: what it costs, how long until first signal, how it is measured, and which parts N8N + a local model can automate. Every automation must name where its decisions get logged as training data.
+- Research-only: no installs, no config changes, no DB writes, no credentials, no commits. Every claim has a URL and date; write UNVERIFIED rather than guess.
+- Etsy-specific: quote Etsy's current rules on AI-generated content disclosure, the creativity standards (what counts as "made by you"), POD partner disclosure, digital-download rules, and fee schedule (listing, transaction, payment, offsite ads). No review manipulation; Etsy Ads is the paid lever — model it.
+
+## Step-by-step
+1. Etsy rules and fees 2026, official pages only. One turn.
+2. Demand method without paid tools: Etsy search-suggest, competitor listing counts vs sales signals (reviews, "bestseller" badges), seasonality; apply to at least 8 product ideas across POD and digital; shortlist 3 with a POD-vs-digital verdict (margin, refund risk, fulfilment effort). Three turns.
+3. Marketing plan: Etsy SEO (titles, tags, attributes, photos/mockups, listing count strategy), launch pricing and free-shipping thresholds, Etsy Ads starter budget with expected ROAS, offsite ads opt-out question, and 2 off-Etsy evergreen assets with no persona (Pinterest is allowed ONLY if it can be scheduled hands-off; state whether that is realistic and compliant). Two turns.
+4. Viability: unit economics per product type (Printify base cost, Etsy fees, price, ad cost), time-to-first-sale evidence, kill criteria after 30 listings / 90 days.
+5. Write `research/etsy-business-evaluation.md` with sections: Verdict; Rules and fees that bind us (quotes); Product scoring table + 3 shortlisted products with POD/digital verdict; Marketing plan (SEO, pricing, Ads budget model, off-Etsy assets, 30-day calendar); What N8N/local model automate (keyword scans, listing drafts, sales logging) + training-log table; Kill criteria; Grill-me questions (≤6); Sources. Append one row to `research/INDEX.md`.
+
+## MANDATORY WRAP-UP (required regardless of success or failure)
+
+Before stopping for ANY reason — task complete, error, or approaching turn limit — output this JSON as your final message. Do not stop without it.
+
+{
+  "status": "succeeded|partially_succeeded|failed",
+  "actions_taken": ["action 1 — outcome", "action 2 — outcome"],
+  "actions_failed": ["action — reason it failed"],
+  "notes": "anything relevant for the next session",
+  "viability_verdict": "go|go_with_changes|no_go + one sentence",
+  "first_customer_path": "<one sentence>",
+  "monthly_budget_needed_usd": "<number or 0>"
+}
+
+If you hit --max-turns before finishing, set status="partially_succeeded" and list what remains in actions_failed.
+
+--max-turns 15
+If you issue the same tool call or command twice with identical arguments, STOP immediately and output the mandatory wrap-up with status=partially_succeeded.
@@ -9,8 +9,11 @@ Canonical copies live here; mirrored to Obsidian `Resources/Prompts/`. Each prom
 | A | [A_ai-harness-layer-research.md](A_ai-harness-layer-research.md) | personal_projects 221 | claude-config | `research/ai-harness-layer-research.md` |
 | B | [B_jcode-vs-claude-code.md](B_jcode-vs-claude-code.md) | personal_projects 239 | claude-config | `research/jcode-vs-claude-code.md` |
 | C | [C_hyperagent-marketing-fit.md](C_hyperagent-marketing-fit.md) | business_projects 41 | `/opt/appdata/docker/Business/API Idea` conversation | `research/hyperagent-marketing-fit.md` (copy to Obsidian `Business/Research/`) |
+| D | [D_api-business-evaluation-and-marketing.md](D_api-business-evaluation-and-marketing.md) | business_projects 44 | claude-config or API Idea | `research/api-business-evaluation.md` |
+| E | [E_kdp-business-evaluation-and-marketing.md](E_kdp-business-evaluation-and-marketing.md) | business_projects 45 | claude-config | `research/kdp-business-evaluation.md` |
+| F | [F_etsy-business-evaluation-and-marketing.md](F_etsy-business-evaluation-and-marketing.md) | business_projects 46 | claude-config | `research/etsy-business-evaluation.md` |

-**Spawn rule:** one at a time (A → B → C), `subagent_type: general-purpose`, paste the prompt body
+**Spawn rule:** one at a time (A → B → C done 2026-09-08; next D → E → F, each followed by the `agent-wrap` checklist), `subagent_type: general-purpose`, paste the prompt body
 verbatim. After each: read the report, add the INDEX.md row if the agent missed it, then run the
 follow-up grill-me before the next spawn. Works on Fable 5.1 or Opus 4.8 unchanged.