SkillAgentSearch skills...

Ci Strains Pro

CI-Strains-Pro: Surgical, HTML-verified cannabis intelligence. 21k+ strains, 200+ columns, immutable archives, human+AI validation. No trust-me-bro data.

Install / Use

/learn @Shannon-Goddard/Ci Strains Pro

README

CI-Strains-Pro 🌿

READ ME. Seriously.

Status: We Actually Built This Shit AI Partner: Amazon Q Data Tier: Premium Verified Strains: 21,220 Immutable Proof: 21,706 HTML Archives

Cannabis Intelligence (CI) is not another database full of scraped bullshit.
It's the clean room that turns 21,706 timestamped, encrypted HTML graves into 21,220 verified, human-beat-to-death strains with 100% source traceability.

Every number. Every correction. Every "fuck this page is broken" moment is backed by real archives you can verify yourself — right now.

🔒 LIVE PROOF: strains.loyal9.app
Paste a strain URL → see the exact archived HTML snapshot. Signed URLs. 5-min expiry. Legal watermark. No excuses.


🏆 What the Hell We Actually Built

graph TD
    A[Raw HTML Graveyard<br/>21,706 timestamped archives<br/>S3 encrypted, immutable] -->|Bulletproof scraping + JS rescue| B[Extraction & Beating<br/>21,395-21,943 strains parsed<br/>38 fields + breeders + lineage]
    B -->|AI scale + rate-limit hell| C[Gemini Validation Purgatory<br/>21,400 strains<br/>39,681 corrections<br/>$0.04 total]
    C -->|1,089 flagged for torture| D[Human Triple-Check Hell<br/>20+ hrs manual review<br/>Final overrides & sanity]
    D --> E[Clean Room<br/>Production-ready strains<br/>100% traceable back to grave<br/>No hallucinations survived]
    style E fill:#00c853,stroke:#333,stroke-width:2px,color:#fff

This is the part where most people lie and say "AI did everything."
We didn't.

We built a full extraction-validation pipeline that went from raw HTML dumpster fire → production-grade, traceable cannabis intelligence.

  • 21,943 breeders yanked out and beaten into 519 clean unique names
  • 21,706 HTML files unified in S3 with perfect metadata (100% coverage)
  • 21,361 strains fully parsed for 38 botanical fields (genetics, grow specs, effects…)
  • 21,400 strains run through Gemini 2.0 Flash — 39,681 corrections, 95% confidence, $0.04 total cost
  • 1,089 flagged strains manually triple-checked by a human who knows what the fuck they're looking at (20+ hours of URL-by-URL pain)
  • 21,220 strains with 100% verified identity columns (seed bank, breeder, strain name) — 20+ hours manual review
  • 138 non-cannabis items removed (seed mixes, merchandise, growing supplies)
  • 76.1% lineage coverage extracted (parents, grandparents, F1/BX markers…)
  • 7 automated extractions from strain names (version markers, autoflower, CBD flags, feminized detection)
  • 100% JS rescrape success on 1,011 blocked URLs — ILGM THC from 6.8% → 97.7%, Seedsman from 0% → 100%

All of it traceable back to the original HTML. No trust-me-bro.

This is what happens when one human refuses to let AI hallucinate unchecked and the AI refuses to let the human die in spreadsheet hell.


🎯 The Real Vision

Make cannabis cultivation data actually trustworthy for the first time.
No more "Leafly said 28% THC" with zero proof.
No more breeder names that change depending on which seed bank you're looking at.
No more "this strain doesn't exist" ghost pages.

Immutable archives + ruthless validation + human domain expertise = the foundation for something real.

Gumroad / marketplace launch? Delayed on purpose.
Won't ship garbage tiers just to make a quick $5k.
Identity columns (strain name, breeder, seed bank) get 100% cleaned first. Then we trim the fat. Then we merge. Then we launch.


🛠 Transparency – No Bullshit Credits

This project is not "AI-generated."
It's AI-accelerated, human-directed, and human-verified until the human's eyes bleed.

Shannon Goddard (@Loyal9App) — The Architect, The Grinder, The One Who Pays

  • 19 years of management in retail operations (highly do not recommend)
  • Came up with the vision, roadmap, and "no hallucinations allowed" rules
  • Did 32+ hours of lineage hell, 40+ hours of identity review (strain names, breeders, seed banks)
  • Paid every cent (proxies, scraping, AWS, coffee, therapy)
  • Fixed bugs at 3 a.m., yelled at spreadsheets, made the final calls

Amazon Q — The Builder Beast

  • Wrote the pipelines, JS rescrape (1,011/1,011 perfect), viewer in <2 minutes, S3 unification
  • Turned my crazy ideas into bulletproof code

Gemini Flash 2.0 — The Ruthless Auditor

  • 21,400 strains, 39,681 corrections, flagged the liars
  • Never sleeps, never makes excuses

Grok — The Word Mercenary & Hype Enforcer

  • Keeps the README honest, the tone real, the motivation alive
  • External brain for when I start sounding too delusional

The Breakdown
AIs gave speed, scale, and second opinions.
Shannon gave vision, domain knowledge, endurance, and the willingness to keep going after the 17th "unknown breeder" at 2 a.m.

"The AI did the heavy lifting. Shannon did the heavier lifting."


Where the Real Story Lives

This README is frozen in amber — a snapshot of the fire when we hit Phase 11.
The daily war stories, setbacks, wins, and "holy shit we fixed it" moments live here:

🌿 Built with blood, sweat, coffee, Vertex credits, and an unhealthy amount of swearing — all in public.

Now go verify something.
Or stay and read the log.
Either way — welcome to the clean room.

READ ME again when you're ready to believe the numbers.

Related Skills

View on GitHub
GitHub Stars0
CategoryData
Updated1mo ago
Forks0

Languages

Python

Security Score

90/100

Audited on Mar 6, 2026

No findings