| name | Hilary Gridley — Judgment Encoding & Anti-Slop Management OS |
| description | Stop the slop upstream: turn tacit expert judgment into narrow deployable evaluator tools (edit pairs → patterns → plain-English rubric → system prompt), articulate what-good-looks-like at every altitude, redesign workflows AI-native (nothing-is-a-surprise, one-step-further), and build the Taste Profile — the canonical context asset that raises the floor of all AI and human work. From the ex-Whoop team lead behind 'How to Be a Super Manager with AI.' |
| version | 1.0 |
| workflows | 13 |
| source | Marketing Against the Grain deep-dive (8,716 words + slide capture, 2026-07-28 forge) |
Hilary Gridley — Judgment Encoding & Anti-Slop Management OS
Everyone else detects slop after it's produced. Gridley prevents it upstream: slop is an unarticulated quality bar, and the cure is a manager's tacit standard made explicit, encoded into narrow tools, and installed as operating culture. Crown jewel: the complete pipeline for turning taste into tooling — edit pairs → pattern mining → criteria → plain-English pass/fail rubric → deployable evaluator. "Anyone has access to your brain."
Load genius.md before any workflow. Deep extraction: extractions/hilary-gridley/extraction.md. Slide capture: extractions/hilary-gridley/visual-context.md.
Available Workflows
Tier 1 — Foundation
| Workflow | Produces | Use When |
|---|
| hg-judgment-encode | A deployable evaluator tool minted from real edit pairs (pattern table + criteria + plain-English rubric + system prompt + validation) | An expert's edits exist and their judgment should scale beyond them — the flagship |
| hg-quality-bar | Quality Bar doc + accountability contract at all three altitudes (time / portfolio / artifact) | A team/role/operator is producing slop and nobody has ever stated what good looks like |
| hg-ai-native-redesign | Two-panel spec: today's friction-mapped workflow vs the proactive AI-native rebuild, time-badged, with build order | Any recurring reactive situation deserves the months-to-minutes treatment |
| hg-slop-diagnostic | Root diagnosis (three roots + flywheel direction + leaking layer) + ordered intervention prescription | Slop is visible but the cause isn't — run before prescribing any tool |
Tier 2 — Practitioner
| Workflow | Produces | Use When |
|---|
| hg-evaluator-fleet | Scored tool inventory + wave-1 tool cards (max 3), each artifact × audience × outcome | Planning which evaluators to build, in what order — the anti-second-brain gate |
| hg-edit-pair-harvest | Structured Column A/B corpus with provenance grade (or a collection brief) | Encoding is wanted but the evidence base isn't assembled yet |
| hg-feedback-script | Speakable graduated-feedback script + escalation move | Slop just landed on your desk and the response must build iteration, not defensiveness |
| hg-kick-the-crutch | Crutch/assistant/coach verdict + redesign spec + flywheel check | Any existing AI tool might be making its users worse |
| hg-exec-editor | The get-to-yes editor for high-stakes messages (rubric + system prompt + validation) | Upward/outward comms carry asymmetric downside — her canonical tool, instantiated |
Taste Profile Cluster (the moat asset)
| Workflow | Produces | Use When |
|---|
| hg-taste-profile | The canonical three-layer Taste Profile (emotional customer understanding + brand narrative canon + quality bar), provenance-grounded | A brand/operator needs the one context asset that upgrades all AI and human output — spec: references/taste-profile-spec.md |
| hg-taste-profile-offer | Productized engagement package: positioning, deliverable stack, proof arc, 10-day delivery, send artifacts | Selling the Taste Profile build as a client engagement (send-before-build binds) |
Tier 3 — Stacking
| Workflow | Produces | Use When |
|---|
| hg-verdict-to-evaluator | Harness-native evaluators mined from Farrice's own verdict/edit logs + "new since last codification" pattern report | The sitting Column A/B data in this system should become fireable quality gates |
| hg-surprise-audit | Surprise inventory + proactive signal design table + wire-this-week build notes | Any operation still runs on humans noticing things |
Stacking Guide
| Pair with | For |
|---|
voice-os / voice-ratchet / felt-verdict-capture | hg-verdict-to-evaluator consumes their logs; minted evaluators cite VOICE-CARD as canon |
context-profile-architect | Taste Profile (content layer) → 01-architect (machine-native JSON twin); 02-excavate feeds Layer 1 depth |
avatar-machine / icp-deep-canvasser | Phase-0 GROUND discipline for Taste Profile Layer 1; the profile is the anti-ICP complement, not a replacement |
geoff-woods-ai-thought-partner | Bar-setting ritual and taste interviews run as co-creation (/gw-*) |
wargame-os | Two halves of judgment transfer: failure-maps for executors + rubrics for evaluators |
nate-b-jones (harness design) | hg-surprise-audit + hg-ai-native-redesign feed harness audits; sniff-check ↔ evaluator criteria |
oren-anti-slop-classifier / anti-slop-audit / prose_classifier | Output-side detectors as the LAST line; this skill builds the upstream bar they backstop |
godin-remarkability-engine | The remarkability premium (slop world rewards OMG work) as offer/positioning fuel |
/offer-redteam + offer_gate | Mandatory pre-launch pass for hg-taste-profile-offer |
Quick Reference
- Genius Context: genius.md — load first, always
- References: taste-profile-spec.md · loop-frameworks.md (Bodnar/Flanagan material, attributed) · source-quotes.md
- Default entries: slop somewhere →
hg-slop-diagnostic. Scale someone's judgment → hg-judgment-encode. One asset to raise a brand's floor → hg-taste-profile.
- The three-question spine of the whole system: What does good look like? Where's the evidence of your taste? What if the AI went one step further?
Execution Prompts (structure-pure v2)
10 deterministic practitioner prompts — each carries an Output Contract, Output Skeleton, and Quality Gate. When a deliverable matches one, Read it and honor its contract instead of improvising the output shape.
- Hilary Gridley — AI-Native Two-Panel Redesign —
skills/hilary-gridley/references/prompts-v2/ai-native-redesign.md
- Hilary Gridley — Edit-Pair Corpus Assembly —
skills/hilary-gridley/references/prompts-v2/edit-pair-harvest.md
- Hilary Gridley — Evaluator Fleet Plan —
skills/hilary-gridley/references/prompts-v2/evaluator-fleet.md
- Hilary Gridley — Executive Editor (Get-to-Yes) —
skills/hilary-gridley/references/prompts-v2/exec-editor.md
- Hilary Gridley — Graduated Feedback Script —
skills/hilary-gridley/references/prompts-v2/feedback-script.md
- Hilary Gridley — Evaluator Tool from Edit Pairs —
skills/hilary-gridley/references/prompts-v2/judgment-encode.md
- Hilary Gridley — Quality Bar & Accountability Contract —
skills/hilary-gridley/references/prompts-v2/quality-bar.md
- Hilary Gridley — Slop Root Diagnosis —
skills/hilary-gridley/references/prompts-v2/slop-diagnostic.md
- Hilary Gridley — Taste Profile Productized Offer Package —
skills/hilary-gridley/references/prompts-v2/taste-profile-offer.md
- Hilary Gridley — Taste Profile (Three-Layer Context Asset) —
skills/hilary-gridley/references/prompts-v2/taste-profile.md
