Paid cast
Groundledger: MIT agent skills for Claude Code, Cursor & Cowork — grounded blog SEO/AEO, PROBE site SEO, SAFE paid media, SPARK editorial, SNAP social, ORBIT discovery
npx -y skills add dranshrad/groundledger --skill paid-castAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 12 days oldThe repository was created 12 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Groundledger paid-media skill with SAFE defaults (Snapshot → Advise → Fence → Experiment). Observe-only unless MutationLatch opens. Modes: audit, plan, math, brand, budget, trial, attrib, landing, creative, pace, optimize-draft. Use for Google/Meta/LinkedIn/Microsoft-style account reviews, unit economics, brand safety lattices, experiment decks, and draft change packs — not live spend edits without explicit approval. Claude Code, Cursor, Cowork.
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
4.3 KB, 917 tokens by cl100k_base, as published. Nobody here has run it
Paid Cast (SAFE media)
Groundledger paid acquisition desk. Complements orbit-discovery (organic surfaces),
site-signal (landing SEO), and social-cast (organic social).
Default: observe-only. No platform mutations, no budget edits, no status flips unless MutationLatch is open.
Grounding: ../blog-engine/references/grounding.md — no invented ROAS,
CPAs, or “industry benchmarks” without a cited source the user accepts.
SAFE loop
| Step | Name | Rule |
|---|---|---|
| S | Snapshot | Read exports / UI dumps; claim + evidence + confidence |
| A | Advise | Ranked actions with unit-econ impact hypotheses |
| F | Fence | Ceilings: budget delta, geo, audience, creative, date |
| E | Experiment | Prefer trials over permanent flips |
MutationLatch (writes)
Latch stays closed until all are true:
- User explicitly asks to draft or apply a change pack
- Scope listed (accounts / campaigns / ad sets)
- Ceilings set (max daily $, max % budget move, kill date)
- Idempotency key or rollback cue recorded
- User says approve on the exact draft
Without latch: deliver audit + draft-only recommendations.
Never pretend a change was applied in-platform.
Claim / evidence / confidence
Every material finding:
Claim: …
Evidence: … (export row, screenshot note, UI path)
Confidence: high | medium | low
Low confidence → ask or mark unknown. Do not bluff attribution.
Modes
| Mode | Job |
|---|---|
audit | Account / campaign health Snapshot |
plan | Channel mix + phased media plan |
math | Unit economics (CAC, payback, contribution) |
brand | Brand Lattice (claims, exclusions, tone) |
budget | Budget Lattice (caps, pacing, reserves) |
trial | Experiment deck (hypotheses, metrics, stop rules) |
attrib | Attribution comparability gate |
landing | Post-click landing audit (pairs with site-signal) |
creative | Fatigue / variant gaps |
pace | Delivery / spend pace vs plan |
optimize-draft | Change pack under MutationLatch |
Platform packs (Google / Meta / LinkedIn / Microsoft-style) are draft mutation templates only — never auto-execute. references/mutation-latch.md.
Unit economics (math)
Require inputs the user can supply: AOV/LTV proxy, margin, target CAC or payback.
If missing: compute scenarios labeled assumption, not facts.
references/unit-economics.md.
Brand Lattice (brand)
Allowed claims · banned claims · competitor rules · sensitive topics · offer language. Creative and landing copy must pass the lattice before scale.
Budget Lattice (budget)
Daily/monthly caps · channel reserves · test budget slice · emergency kill.
Pace mode watches spend vs plan without changing bids.
Trial Deck (trial)
Hypothesis · primary metric · guardrails · sample / time stop · rollback.
One primary metric per trial. No “test everything.”
Attribution gate (attrib)
Before comparing channels: same window, same conversion def, same currency, same inclusion rules. If incomparable → say so; do not crown a winner.
Landing (landing)
Message match, offer clarity, speed honesty, form friction, proof proximity.
Deep SEO issues → hand off to site-signal. Copy rewrites → blog-engine / editorial-pass.
Creative (creative)
Fatigue signals from user data only. Variant matrix: angle × format × offer.
Respect Brand Lattice.
Context files
BRAND.md · ORBIT.md · optional MEDIA_LATCH.md (ceilings + approvals).
Bridges
- Landing SEO →
site-signal - Organic social →
social-cast - Long-form proof pages →
blog-engine - Surface strategy →
orbit-discovery