Sun 31 May 2026
Fifth pass — 2026-05-31 (agnt.one — generic RAG-chatbot-widget SaaS)
One item this pass. Surfaced by Anthony 2026-05-31 from a feed-skim of agnt.one. Gavin Bintz (@bintz_gavin). No-code ChatGPT-backed chat widget embedded on a website. Trains on URLs + uploaded docs. Embed widget or standalone subdomain. $8 / $50 / $150 per month. US SaaS, OpenAI backbone, Vercel-hosted. Sebastian scored without deep evaluation — fails three Articulate filters at once.
| Item | Source | P | Q | S | Total | Move |
|---|---|---|---|---|---|---|
| Agent One (agnt.one) — no-code ChatGPT-powered chat widget; trains on URLs + docs; embed or standalone subdomain; $8–$150/mo. Built by Gavin Bintz. US SaaS, OpenAI backbone. | Anthony's feed-skim 2026-05-31 | 0 | 0 | 1 | 1/9 | Skip. Fails the data-sovereign positioning (US SaaS + OpenAI under the hood — wrong substrate for the UAE/KSA regulated mid-market the data-sovereign-stack bundle targets). Consumer-tier monthly pricing ($8–$150) — wrong billing surface for AED 25–80k engagements. Crowded category — Chatbase / SiteGPT / Chaindesk / Voiceflow / Intercom Fin / Lemonade-AI already occupy this space. If a client asks "we want a chat widget on our site," the answer is the n8n + Postgres + self-hosted LLM gateway pattern inside the data-sovereign-stack, not reselling agnt.one. |
Generalisable filter (new — apply at scoring time)
Any item that is all three of (a) US SaaS reselling OpenAI under the hood, (b) consumer/prosumer tier monthly pricing ($5–$200), (c) in a crowded RAG-chatbot / widget / agent-builder category — defaults to skip without scoring, unless one of the following overrides fires:
- explicit UAE-PDPL / data-residency angle (self-host option, EU/MEA region, signed BAA-equivalent),
- a structural moat we cannot replicate (proprietary eval set, specialised model, exclusive distribution),
- a positioning hook Articulate could steal without reselling the product (e.g. agnt.one's "deploy as standalone SEO-indexed page" trick is interesting — doesn't need their product to execute).
This filter saves analysis cycles on a category that produces a new entrant most weeks. Logged as the RAG-widget-SaaS skip filter.
Re-score triggers
- (a) agnt.one ships a self-host or sovereign-cloud SKU with UAE/KSA region option → re-score S, possibly to 2.
- (b) A specific client engagement scopes "deploy a chat widget on a marketing site" as a fixed-scope discrete deliverable AND the client refuses the data-sovereign build path → re-score S as a reseller margin play.
- (c) Vercel ships first-party RAG-widget primitives that change the build vs buy economics for the data-sovereign bundle.
Sixth pass — 2026-05-31 (Mike-Shafer-style 10-prompt "revenue workflow" pack)
One item this pass. Surfaced by Anthony 2026-05-31 from a viral X post — ten "highest-leverage Claude Code dynamic workflow prompts to generate more revenue, assuming unlimited tokens." Operator-deck shape: Revenue Command Center · Stalled Revenue Recovery · Client Expansion Engine · Churn Prevention · SEO/AI Search Revenue Gap · Offer Rebuild · Sales Call Intelligence · Paid Acquisition Profit Audit · Founder Content-to-Revenue · Automation ROI Hunter. Closes with a "monster meta-prompt" that fan-outs specialist subagents and synthesises into a ranked opportunity model.
| Item | Source | P | Q | S | Total | Move |
|---|---|---|---|---|---|---|
| 10-prompt "revenue workflow" pack — operator-deck of fan-out + rank + execute prompts for revenue surfaces. "Use a workflow" framing; no SKILL.md, no eval gate, no subagent wiring shipped — prompts only. | Anthony X-feed forward 2026-05-31 | 2 | 2 | 2 | 6/9 | Act on as INPUT to existing Articulate canon, not as install. Six of the ten map directly onto MEP delivery + Connie-for-IC pipeline + brand-voice / discover-brand plugins already in our stack. Three are stronger than what we currently scaffold (Stalled Revenue Recovery, Client Expansion Engine, Sales Call Intelligence). Build the superior Articulate versions inside the relevant playbooks. Do not ship the raw prompts as a "Mike Shafer pack" — that would replay the Corey Haines mistake (second pass): install the viral artefact instead of building the productised offer. |
Where each of the ten lands inside our existing stack
| # | Pack prompt | Lands in |
|---|---|---|
| 1 | Revenue Command Center | MEP — already the spine of the Marketing Engine Pilot. Pack's fan-out language is sharper than current MEP brief; lift the subagent list (pipeline, retention, upsells, SEO, paid, outbound, partnerships, ops bottlenecks) into Articulate/offer.md as the canonical MEP surface set. |
| 2 | Stalled Revenue Recovery | Net new — no current Articulate playbook covers 180-day stalled-deal recovery as a discrete deliverable. Productise as Articulate/offers/stalled-revenue-recovery/ — fixed-scope $3–8k engagement, paid pilot, sold as bolt-on to MEP or stand-alone to operators with existing CRM history. Highest-S piece in the pack. |
| 3 | Client Expansion Engine | Net new — productise as Articulate/offers/expansion-engine/. Slots into MEP delivery #6 or as stand-alone retainer-extension play. Mirrors what Connie does for IC; this is the commercial version. |
| 4 | Churn Prevention + Save Plan | Slots into Connie's analyst workflow + the Hermes Agent Curator. Build as a connie/playbooks/churn-save.md subcommand, not a separate offer. |
| 5 | SEO / AI Search Revenue Gap | Already partially covered by the marketing:seo-audit skill installed in our stack. Extend that skill with the AI-search-visibility axis (covered nowhere in current SEO skills) — that's the genuinely new piece. |
| 6 | Offer Rebuild | Maps to Articulate brand + offer workflow Will + Belinda already own. Lift the "3 variants — safest / aggressive / premium" structure into the standard offer-doc template. Small but useful canon edit. |
| 7 | Sales Call Intelligence | Net new at depth. brand-voice:discover-brand covers transcript-as-input. This goes further — buyer pains, objection map, weak-rep moments, missed discovery questions, re-engageable prospects. Productise as Articulate/offers/sales-call-intelligence/ — paid Gong/Granola transcript audit, $2–4k fixed scope. |
| 8 | Paid Acquisition Profit Audit | Inside Supermetrics + the planned data-query playbook. Build as a Sebastian playbook SuperSebastian/playbooks/paid-audit.md. |
| 9 | Founder Content-to-Revenue Engine | Maps to the LinkedIn content engine Will runs + the Articulate/linkedin-content/ folder. Lift the "ideas → posts/podcasts/YT/newsletter/lead-magnets/CTAs with buyer-intent scoring" framing into Will's planning doc. |
| 10 | Automation ROI Hunter | Sebastian-native territory. Build as SuperSebastian/playbooks/automation-roi-hunter.md. Pair with the n8n install already on the radar (first pass 8/9). |
Why P=2 (not 3)
The pack is prompt-shaped, not skill-shaped. To actually execute it you need: (a) the data surfaces it asks for (CRM exports, transcripts, ad data, analytics), (b) a subagent fleet wired to fan out the work, (c) eval gates so the "execute safe implementation work in parallel" line doesn't ship slop. We have most of (a) and (b) already — Sebastian, Connie, Will, Belinda, the brand-voice plugin, n8n, Higgsfield, Supermetrics. We do not have (c) at the level the meta-prompt implies. P promotes to 3 once the productised offers (Stalled Revenue Recovery + Client Expansion + Sales Call Intelligence) ship as scaffolded playbooks and one runs against real client data.
Why Q=2
The fan-out shape (Revenue Command Center subagents) is sharper than current MEP brief language and lifts cleanly into Articulate/offer.md. The "3 variants — safest/aggressive/premium" framing for offer rebuilds is a clean canon-edit. The "ranked opportunity model with estimated upside, confidence, owner, next action, required assets, 7-day execution plan" rubric is the right shape for every diagnostic deliverable Articulate ships — codify it as the standing rubric for MEP discovery output.
But: the prompts ship without eval gates, without benchmark scoring, without provenance discipline. Same failure-class as the SEB-gate four-strike (Rule 17) — vibe-mode "use a workflow with unlimited-token depth" is exactly the framing that produces confident-shaped slop. Class-match against the mymentor carousel (wrong-tool fail) and Takis four-brand cycle (brand-spec drift). Q caps at 2 until paired with the SkillOpt benchmark-gated loop (fourth pass) or the Taste Pipeline rule-set discipline (third pass).
Why S=2
Three of the ten (Stalled Revenue Recovery · Client Expansion Engine · Sales Call Intelligence) productise into discrete fixed-scope Articulate offers at $2–8k each. Real S=3 territory at the offer level, but the pack itself is the input, not the product. S=2 holds for the pack. S promotes to 3 on a per-offer basis when each one signs its first paying engagement.
Class-match against failures.md
- 2026-05-24 SEB-gate four-strike — meta-prompt's "use a workflow with unlimited-token depth. Do not optimize for brevity" is the same vibe-mode framing that triggers Sebastian's auto-fire failures. Adopt the structure (fan-out → rank → execute → verify); reject the framing (token-uncapped, brevity-optimised-against).
- 2026-05-20 Corey Haines marketing-skills (second pass) — viral operator-deck of prompts. Same move applies: build the superior Articulate version inside our offer surface, do not install the viral artefact.
- 2026-05-18 Takis four-brand-cycle — fan-out subagent execution without brand-anchor enforcement reproduces the four-brand drift. Any pack-derived workflow must route through Dylan's
Design/PromptLibrary/<brand>/anchors.mdbefore generating any visual artefact.
What's actually new (vs already in canon)
Most of the pack restates what Articulate already runs. Three pieces are genuinely additive:
- Stalled Revenue Recovery as a discrete diagnostic — currently nowhere in the offer stack.
- Buyer-intent scoring as a content-engine input (prompt 9) — Will runs the engine without explicit intent-scoring layer; this is a clean addition.
- AI-search visibility axis on SEO (prompt 5) —
marketing:seo-auditdoes not cover this. Net new sub-skill.
The rest is reframing, sharpening, or re-statement of canon.
Re-score triggers
- (a) Stalled Revenue Recovery offer ships as
Articulate/offers/stalled-revenue-recovery/with first paid pilot scoped → S promotes to 3, total 7/9. - (b) Sales Call Intelligence offer ships with first Gong/Granola audit billed → S promotes to 3, total 7/9.
- (c) All three new offers run paired with SkillOpt benchmark-gate (fourth pass) → Q promotes to 3, total 7-8/9.
- (d) Meta-prompt's "unlimited token depth + no brevity" framing ships into any live workflow without SEB-gate → failures.md negative, item drops to skip.
Build path
Three offer scaffolds to spin up (in order of S-impact):
Articulate/offers/stalled-revenue-recovery/— folder + offer.md + delivery.md + pricing.mdArticulate/offers/sales-call-intelligence/— folder + offer.md + delivery.md + pricing.mdArticulate/offers/expansion-engine/— folder + offer.md + delivery.md + pricing.md
Plus three canon edits:
Articulate/offer.md— lift the Revenue Command Center subagent surface set into MEP descriptionArticulate/linkedin-content/planning doc — add buyer-intent scoring layer per prompt 9SuperSebastian/playbooks/—paid-audit.md+automation-roi-hunter.md
Not built this pass — scaffold sequence handed to next planning session. Rick gates the per-offer scope before any build.