Files
directory-engine/README.md
T

61 lines
3.3 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# Directory Engine
**Leonard-powered niche directory business** (inspired by [Startup Ideas Podcast](https://x.com/startupideaspod/status/2091238981337858473)).
A directory is curated data: "here are the best X in this topic." Built for AI agents. Research agent produces high-quality Markdown pages daily; Leonard (engineer) publishes them. Later: affiliate collection + outbound deal agents. Newsletter on subdomain for sponsors.
Target: $10K–$100K/month once traffic compounds. SEO shift favors curated sources for ChatGPT / Google AI Overviews.
## Core Playbook (for Leonard / Hermes)
1. **Pick the niche** — something with taste (e.g. local AI tools & LLM stacks, El Dorado County verified service providers, party board games, pickleball gear). Record decision in `decisions.md`.
2. **Site foundation** — Astro (excellent for static directories). Structure under `src/content/` or `pages/` as Markdown collections.
3. **Research agent** — Daily: pick next brand/item → investigate (web, X, reviews, pricing, differentiators) → write polished Markdown page with frontmatter (title, slug, tags, affiliate-ready fields, evidence).
4. **Engineer (Leonard)** — Receives Markdown → validates → commits to this repo (or site repo) → triggers build/deploy if configured.
5. **Rinse & repeat** until thousands of monthly visits.
6. **Only after traffic**: agent that harvests referral/affiliate links + does outbound for deals. Warm traffic → newsletter sponsors.
**Rule**: Never solicit deals before traffic exists.
## Repo Layout
```
directory-engine/
├── README.md # this file
├── decisions.md # niche choice, rationale, dates
├── status.md # current state, next actions for Leonard
├── agents/
│ ├── research-prompt.md # system prompt + workflow for research agent
│ ├── engineer-prompt.md # Leonard publish checklist
│ └── outbound-prompt.md # future: referral + deals (post-traffic)
├── templates/
│ └── page.md # Markdown page template with frontmatter
├── pages/ # generated / curated pages (or symlink to Astro content)
│ └── _index.md
├── src/ # (future) Astro site source
└── scripts/ # optional helpers
```
## Leonard Operating Instructions
- Read `status.md` and `decisions.md` first on every run.
- Use research agent prompt to produce one new page per cycle.
- Commit with clear message: `feat(page): add <slug> — <one-line summary>`.
- After commit, update `status.md` with what was done and next candidate.
- Prefer verified facts + sources. No hallucinated claims.
- Keep pages useful for both humans and AI overviews (clear structure, lists, comparisons).
## Next Actions (initial)
1. Decide niche → write `decisions.md`.
2. Flesh out `agents/research-prompt.md` and `templates/page.md`.
3. Seed 3–5 initial pages manually or via first agent run.
4. Stand up minimal Astro site (or pure Markdown + static generator) and point deploy.
5. Wire daily cron / Hermes schedule for research → Leonard publish loop.
## Links
- Source idea: https://x.com/startupideaspod/status/2091238981337858473
- Gitea: https://gitea.vps1.afterthedemo.com/Tony_tech/directory-engine
- Related: Tony_tech/veripath, Tony_tech/personal-agent-os, Tony_tech/agent-outputs