* feat(product-marketing): version history for the context document (2.1.0)
.agents/product-marketing.md now carries a Document version header + a
newest-first Changelog section. The update flow reads the version/recent
changelog on open and, on save, bumps the version + prepends an entry
naming what changed and why (never rewriting past entries). Makes
positioning/ICP changes traceable across a project since every skill
reads this doc. New eval. Repo release 2.8.12. Closes #462... (pmm issue).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(product-marketing): clarify typo-fix exception (no version bump or changelog) + ISO dates, per codex
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(video): reverse-engineer a viral edit into a reusable edit spec (2.1.0)
New references/edit-anatomy.md: decompose a reference edit into a beat
sheet (per-beat table + 3-5 signature moves) across nine edit-anatomy
dimensions, pulled via watch-video/social-fetch, reviewed once, then
executed in Remotion/CapCut/an AI editor. The tool-agnostic decomposition
half of 'copy any viral edit' (Arcads' Omni generation is proprietary).
Originality guardrail: copy editing grammar, never the reference's
footage/script/music. SKILL.md workflow + triggers + eval. Repo release
2.8.11. Closes#456.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(video): soften edit-style IP claim per codex — techniques usually reusable, finished-work reproduction still risky (§102(b)); add SFX/samples
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(ad-creative): add AirDrop as a 4th iOS-native reveal ad surface (2.8.0)
Extends references/imessage-video-ads.md with an AirDrop Reveal surface:
the incoming-AirDrop-card hook + accept-tap-as-reveal interaction, the
transfer progress-ring signature motion, sender-name casting, the
AirDrop swoosh sound, receiver/sender POV, and the family grounding
rules. H1/intro/surface-table/pointer updated to four surfaces; new
'AirDrop ad' trigger; new eval. Format popularized by Shiv Sakhuja /
Gooseworks; original craft (their recipe is private). Repo release
2.8.10. Closes#452.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(ad-creative): AirDrop UI fidelity per codex — gray Decline (not red), valid sender names
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(ad-creative): shareable creative review page (client approval artifact) (2.7.0)
Self-contained HTML template (assets/creative-review-template.html) that
presents generated concepts as in-feed platform mockups with a frame-by-
frame storyboard, selectable copy variations, and a required grounding
disclosure — the visual upgrade to INDEX.md. Driven entirely by a DATA
object; no build or deps. Full data model + grounding rules in
references/creative-review-page.md. Repo release 2.8.8. Closes#440.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(ad-creative): harden review-page template per codex review
- P1: DATA moved to inert <script type=application/json> block (values
can't execute) + documented \u003c escaping rule for the </script>
breakout; malformed JSON now shows a clear error banner
- P1: platform list whitelisted to instagram/facebook + escaped in
attr/label (was raw-interpolated)
- platform toggle now renders real FB chrome (link card + text actions)
vs IG (swipe bar + icon actions) — was visually inert
- broken/missing images fall back to the label+prompt placeholder as
documented (placeholder base layer + image removes itself on error)
- doc consistency: agency wording, JSON data-model note, VERSIONS trigger list
Verified in-browser: FB/IG toggle, error banner, and HTML-payload escaping.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
The CLI no longer unconditionally symlinks into .claude/skills — Claude Code
gets .claude/skills only when selected (interactively or via -a claude-code).
Agent-detected non-interactive runs can land universal-only in .agents/skills,
which Claude Code does not read. Verified empirically against skills CLI 1.5.15
and Claude Code 2.1.173.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(ai-seo): citations vs recommendations — AI visibility ladder + listicle risk (2.2.0)
Adds references/citations-vs-recommendations.md: the visibility ladder
(retrieved/cited/mentioned/recommended), the self-promotional listicle
risk for emerging brands, what earns recommendations (offsite
consensus), recommendation value data, and the attribution blind spot.
Caveats wired into content-patterns.md, content-types.md, and the
competitors skill (2.0.1). Repo release 2.8.6. Closes#427.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(ai-seo): correct stat framings per codex primary-source check
69% is per-citation (224/323), not per-query; Scrunch figures split by
metric and framed observationally; SimilarWeb attribution phrasing
corrected; 'reliably' softened to 'often' in competitors.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(ai-seo): residual stat wording in VERSIONS.md and eval expectation
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Four token-driven styles (monoline editorial, Swiss typographic, wireglow,
duotone screenprint) driven by a brand-slots contract (FIELD/INK/ACCENT/
TYPE FEEL + per-brand constraints) so any company's brand system drops in.
Per-style motion notes. All four verified with a real brand's tokens.
Repo release 2.8.3.
* feat(ad-creative): faceless motion-style video ad format (2.3.0)
New references/motion-video-ads.md: fully generated 15-45s concept videos
(styled poster stills -> image-to-video living motion -> TTS -> word-timed
captions, ~$3-6 per video). Provider-agnostic pipeline with one-key Gemini
path, five-style visual library with fill-in prompt formulas, motion prompt
formula, and production QC gotchas earned shipping the format end-to-end.
Format credit: @borjafat / Bomx super-video-maker (re-expressed, extended).
Repo release 2.8.2.
* fix: preserve json formatting in version bump (no unicode escapes, keep newline)
Repo x.y.z semantics: x = repo-wide changes, y = new skills, z = updates
to existing skills. The iMessage ad format is an update to ad-creative,
so it ships as 2.8.1, not 2.9.0. Convention now documented in AGENTS.md.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Covers 2.7.0 (ads 2.2.0 B2B expansion, ad-creative 2.1.0 static
templates, marketing-loops 1.1.0), 2.8.0 (marketing-council), and 2.9.0
(ad-creative 2.2.0 iMessage chat-reveal video ads). plugin.json and
marketplace.json were still at 2.6.0 — 2.7/2.8 shipped without bumping.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(ad-creative): add iMessage chat-reveal video ad format (2.2.0)
Distills the Gooseworks iMessage video ad format (credited; source repo
is unlicensed so content is re-expressed, not copied) into a new
reference: six concept angles, script/pacing rules, three production
routes, craft rules, grounding/compliance guidance, and a ship
checklist. Wires it into SKILL.md with new trigger phrases and adds an
eval. Closes#411.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(ad-creative): address codex review of imessage-video-ads reference
Reword the Gooseworks route (public but unlicensed repo, agent-agnostic
invocation), replace the "eavesdropping on a real recommendation"
framing with dramatization + paid-placement disclosure guidance, add an
Apple trade-dress legal-review note with a generic-chat fallback, and
tighten the technical details (recordVideo.size, xfade input matching,
limiter ceiling under full scale).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix: restore trailing newline in ad-creative evals.json
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat: add marketing-council skill — simulated board of advisors (closes#391)
Twelve research-grounded advisor dossiers with deliberate lens conflicts, three session modes, designated-dissenter seating, disagreement maps, hard anti-fabrication grounding rules, custom advisor support, and evals. Research pass verified framework attributions and flagged misattributions.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix: address codex review — source flagged quotes, correct Sutherland AI positions
Codex found verified Sutherland AI positions (Spectator Jan 2024, MAD//Masters May 2026 via PPC Land) contradicting the dossier's do-not-attribute flag; replaced with the documented positions. Sourced or paraphrased quote-style lines in Ogilvy, Handley, and Dunford dossiers. Reworked eval 5 around recency-routing instead of the false premise.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix: source or paraphrase remaining quote-style lines (codex round 2)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(ads): B2B operations expansion — 5 new playbook references + routing table (2.2.0)
Adapted and re-expressed from practitioner playbooks (notably Ivan Falco's ads-skills; Commons Clause-licensed, so all content rewritten): B2B demand lifecycle + kill rules + lead quality scoring, TCPL-anchored Meta decision system, LinkedIn bidding/scaling/benchmarks playbook, Google intent-ladder playbook, ABM playbook with cross-channel UTM remarketing. RSA output spec moved to its own reference, bringing SKILL.md back under 500 lines.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix: reconcile threshold inconsistencies from codex review of #403
Close penetration-band gap (25-35% hold band), close fatigue-band gaps (warning extends to critical boundary), scope scaling-ads stricter bar explicitly, tie retire rule to per-type critical band, align generic budget-scaling line with the Meta playbook's never-30%+ rule.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix: drop contradictory 60-day penetration rule; label account-level scale guardrail
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* chore: release v2.6.0 — bump ads to 2.1.0 for Meta Andromeda playbook
The v2.5.0 Sabri Suby / creative-as-targeting rewrite of skills/ads/SKILL.md
(#384, #385) shipped without a version bump, so the repo update check
(VERSIONS.md vs local skill metadata) would not surface it to installed users.
- ads: metadata.version 2.0.1 -> 2.1.0 (skill body only; description unchanged)
- VERSIONS.md: bump ads row + add 2.6.0 Recent Changes entry
- marketplace.json: metadata.version 2.5.1 -> 2.6.0
- plugin.json + README: synced via sync-skills.js
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* feat(marketing-loops): add marketing loops skill library
Adds a new `marketing-loops` skill — a library of ~20 repeatable marketing
loops (recurring agent workflows run on a cadence, each with a trigger, loop
body, self-check, and explicit stop/bail-out). Inspired by ForwardFuture's
Loop Library pattern, scoped to marketing ops. Operational cousin of
marketing-ideas: ideas = what to try once, loops = what to keep doing.
- SKILL.md: eight-part loop anatomy, cadence rule (match frequency to signal
speed), when NOT to loop, scheduling guidance (defers mechanics to loopify /
cron), anti-patterns + banned vocabulary.
- references/loop-catalog.md: full library by AARRR + ongoing ops, each loop
orchestrating existing skills (seo-audit, ads, emails, churn-prevention,
pricing, referrals, ab-testing, analytics, etc.).
- VERSIONS.md / README / marketplace.json / plugin.json synced. Total skills 46.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* feat(marketing-loops): expand to 43 loops + 9-part anatomy from review
Incorporates codex review feedback:
- Adds State/idempotency as a 9th anatomy part (last-run marker, dedupe,
cooldown) and splits check-cadence from action-condition.
- Fixes weekly-review self-contradiction (stop was n/a).
- Tightens high-risk loops: newsjacking veto list + human approval;
ad-fatigue real significance bar; lifecycle-email off opens onto
clicks/complaints/bounces; review-harvest consent/FTC/ToS; pricing +
paywall judged on revenue quality not conversion alone.
- experiment-backlog is now a thin wrapper that defers to ab-testing.
- Expands catalog from 19 to 43 loops for thorough funnel coverage
(paid-search mining, retargeting hygiene, LP regression, deliverability,
internal-linking, pSEO quality, PQL, dunning, expansion, trial-conversion,
review-site mgmt, case studies, analytics-anomaly, tracking-QA,
campaign-postmortem, and more), low redundancy.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* fix(marketing-loops): soften loopify refs (not bundled in this plugin)
Per /review: loopify lives in a separate repo and is not part of the
marketingskills plugin. Removed the hard "see loopify" pointer from the
description and Related Skills; kept it as an optional companion in the
Scheduling section with native Claude Code primitives + cron/manual
fallbacks as the always-available path.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* test(marketing-loops): add evals.json
Seven evals covering loop selection + nine-part anatomy, the disciplined
"start with one loop" framing, high-risk handling (newsjacking veto +
human approval), scheduling guidance, the when-NOT-to-loop pushback on
strategy/creative, the dunning revenue-recovery loop with a real stop
condition, and the experiment-backlog thin-wrapper deferral to ab-testing.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* feat(marketing-loops): add loop-template.md for authoring custom loops
Copy-paste template with the nine-part anatomy, ordered fill-in prompts,
a worked blank-to-filled example (a first-call activation loop), and a
pre-ship checklist. Wired into SKILL.md (new Authoring section) and the
catalog closing note.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* feat(marketing-loops): add orchestration, state, and guardrails references
Closes the three real gaps for a skill about autonomous recurring work:
- loop-orchestration.md: how loops compose into a four-layer system
(sensing -> diagnostic -> action -> learning) + a staged rollout path
(tracking + weekly review first, acquisition later, one at a time).
- loop-state.md: where loop state lives (.agents/loops/<loop>.json),
watermark/dedupe/cooldown/in-flight idempotency patterns, and a run-log
format that doubles as a vanity-loop detector. Makes idempotency real.
- loop-guardrails.md: two-tier action model (autonomous-safe vs gated),
spend/send caps + allowlists, CAN-SPAM/GDPR/FTC/ToS mapped to loops,
PII handling, always-escalate list, kill switch, pre-launch checklist.
All three wired into SKILL.md; changelog updated.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Reconciles the existing Audience Targeting Overview section with the
Modern Meta Playbook (Andromeda) section from PR #384. They were soft-
tensioned: the old table presented interest-stacking as a viable Meta
approach, while the new section calls it out as actively harmful in 2026.
The reframe: audience knowledge is still the highest-leverage work — but
the question is WHERE to apply it. The default 2026 answer is creative-
first, targeting-filters-second. How much that ratio tips toward
creative varies meaningfully by platform.
Changes:
- Renamed section "Audience Targeting Overview" → "Audience Understanding
& Targeting" to reflect the broader scope.
- Replaced the 3-row Platform Strengths table with a 6-row platform-by-
platform table showing the creative-vs-filter ratio for each:
- Meta (post-Andromeda): 80%+ creative
- Google Search: 60% filters (keywords still dominant)
- Google PMax / Demand Gen: 70% creative
- LinkedIn: 60% filters (firmographic identity still high-quality)
- TikTok: 70% creative
- X / Twitter: 50/50
- New "Applying audience knowledge to creative" subsection mapping each
type of identifier (demographic, pain points, hopes, objections,
language, niche) to where in the creative it belongs, with wikilinks
to the relevant Modern Meta Playbook sections (one-keyword hack,
4-component retargeting framework).
- Preserved the existing Key Concepts subsection (lookalikes, retargeting,
exclusions) with wikilinks to the modern retargeting playbook for the
updated tactical guidance.
- New "Common failure mode" callout warning about compensating for weak
creative with hyper-precise targeting.
No content was removed — the existing reference link to
references/audience-targeting.md is preserved.
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Source: Sabri Suby's "Learn 97% of Meta Ads in Under 19 minutes" (King
Kong founder, $300M+ in Meta ad spend managed). Captured in
second-brain vault. Adds 2026-era operational tactics that update the
existing skill for the post-Andromeda algorithm shift.
Four new sections + framework updates:
1. Modern Meta playbook (Andromeda era — 2026+) — new top-level section
covering: creative volume as the constraint, statics > polished video
(Meta's algorithm bias + production cost), creative-is-the-targeting
(broad audience + specific long-form creative beats interest-stacking),
the one-keyword identity-trigger hack, AI variant farming with the
100-people prompt, zombie campaigns, don't-make-ads-look-like-ads,
burner-account technique.
2. Retargeting — added "Retarget with DIFFERENT offers" subsection and
the 4-component retargeting framework (objection-handling ad → proof
carousel → other-offers CBO → value-first audit offer).
3. Landing Page Alignment — new section covering the headline-mirror
trick (mirror winning ad headline verbatim on landing page H1/sub/
lead-in copy for 15-20% conversion lift) and the 3-split-tests-minimum
discipline.
4. Scaling discipline (net cash > ROAS percentage) — new section under
Reporting: net cash flow > ROAS percentage at the business level,
break-even ROAS calculation, the 3-hour founder-led data review
discipline, outbound-calling leads who didn't convert as the
insight-to-creative loop.
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
In v2.5.0 we added cross-references in copywriting, launch, pricing, and
sales-enablement descriptions pointing to the new offers skill, but
left those skills' metadata version + VERSIONS.md rows at 2.0.0. The
repo's update check compares VERSIONS.md against local skill metadata,
so users with installed copies would not be notified of the routing/
discovery change. Caught by codex review.
Patch bumps (not minor) because the change is just adding a cross-ref,
not adding new content - matches the precedent of ai-seo 2.0.0 -> 2.0.1
for the Google guide alignment.
- copywriting: 2.0.0 -> 2.0.1
- launch: 2.0.0 -> 2.0.1
- pricing: 2.0.0 -> 2.0.1
- sales-enablement: 2.0.0 -> 2.0.1
- Repo: 2.5.0 -> 2.5.1
- VERSIONS.md updated with new release section + table rows
validate-skills.sh passes 45/45.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds a new offers skill (1.0.0) covering offer construction — the thing
you actually sell, not the page that sells it. Frames the discipline
around two existing frameworks: the Value Equation (Dream Outcome x
Perceived Likelihood / Time Delay x Effort, originally Hormozi, now
standard across direct response and creator-economy training) and the
six-component anatomy of a complete offer (core + bonuses + guarantee +
scarcity + name + price/payment structure).
Self-scoped explicitly: best for services, agency retainers, courses,
coaching, info products, high-ticket B2B, and direct response. Less
load-bearing for pure self-serve SaaS where pricing does more work.
Files:
- SKILL.md (153 lines) - overview, value equation, anatomy, diagnostic
loop, banned vocabulary, anti-patterns
- references/value-equation.md (134 lines) - four levers deep dive with
scoring prompts and worked example
- references/offer-anatomy.md (203 lines) - six components with a
fractional-CMO before/after
- references/guarantee-design.md (172 lines) - eight guarantee types
with decision tree by business type, refund tolerance, and buyer
sophistication
- references/bonus-stacking.md (150 lines) - bonuses-as-objection-
handlers, the math of stated value, the 4-bonus pattern that works,
failure modes
- references/scarcity-urgency.md (175 lines) - honest scarcity formats,
explicit rejection of fake countdown timers and manufactured FOMO
- references/offer-formats.md (240 lines) - default offer format by
business type
- references/examples.md (215 lines) - six anonymized before/after
worked examples
Cross-references added in pricing, copywriting, launch, and sales-
enablement descriptions pointing to offers for offer construction.
Versions bumped:
- Repo: 2.4.2 -> 2.5.0 (new skill = minor)
- Plugin description: 44 -> 45 skills, called out "offer design"
validate-skills.sh passes 45/45.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds a new reference doc and a brief SKILL.md subsection covering
Google's v0.1 Open Knowledge Format spec, introduced on the Google
Cloud blog on 2026-06-12 and shipped inside Knowledge Catalog.
What changed:
- New references/okf.md (104 lines) with the full breakdown
- Brief OKF subsection in SKILL.md under Machine-Readable Files for
AI Agents, linking to the reference
- Description triggers extended: llms.txt, OKF, Open Knowledge Format,
knowledge bundle, agent-readable site
- Brief OKF callout added to the Google AI Overviews section of
references/platform-ranking-factors.md framing OKF as a "register
early" protocol bet rather than a confirmed ranking signal
- ai-seo skill metadata version 2.0.1 -> 2.1.0
- Repo version 2.4.1 -> 2.4.2 (patch: no new skill, just content)
Honest framing carried through both files: Google built OKF primarily
for data-team catalog metadata sharing. The site-readable-by-agents
repurposing popularized by Suganthan Mohanadasan is a legitimate but
secondary use. Treat OKF adoption as protocol-layer registration like
early schema.org, not as a tactic that moves rankings today.
Verification: confirmed Google Cloud blog post is live (datePublished
2026-06-12), confirmed v0.1 spec at GoogleCloudPlatform/knowledge-
catalog/okf/SPEC.md, confirmed Knowledge Catalog product page, and
confirmed Suganthan's free OKF generator at suganthan.com/okf-
generator. WordPress plugin awaiting wp.org approval at time of
writing.
validate-skills.sh passes 44/44.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds a sentence in the intro paragraph pointing to the new offer at
conversionfactory.co/offers/ai-marketing-training. Placed between the
Swipe Files newsletter mention and the Magister autonomous-agent
mention to keep the "learn from content / learn through training /
hands-off autonomous" flow.
Uses the "dangerously good at AI" framing from the launch tweet.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Corey Haines