| name | biznomad-meta-ads-team |
| description | Biznomad's end-to-end Meta Ads team workflow — runs the 5-phase pipeline (spy → bulk creative → pre-launch scoring → competitor angle extraction → full health audit) across any Biznomad client account (HV, Vicelle, D'Lluxe, Liquid Wizdom, Biznomad internal). Chains the existing /ads-* skills into a single productized agency workflow. Use when user says "run the ads team", "full Meta Ads workflow", "Meta Ads team", "5-phase ads", "agency ads workflow", "Biznomad ads stack", or asks to "go through the full Meta Ads pipeline" for a specific client. Inspired by cindie.zhu's Meta x Claude skills reel.
|
Biznomad Meta Ads Team
Productized agency workflow that replaces a $10k/mo agency retainer with a
5-phase pipeline of Claude Code skills. Reusable across every Biznomad client
running Meta Ads.
Before starting (pre-flight checklist — do not skip)
-
Confirm the target client. Never mix client data (MEMORY.md project
isolation rule). Ask: "Which client account — HV (act_…), Vicelle,
D'Lluxe, Liquid Wizdom, or Biznomad internal?"
-
cd into the client's project root. All sub-skills depend on the
right CLAUDE.md, brand-profile.json, and .mcp.json being in scope.
-
Grep memory for client-specific rules — MANDATORY before any
creative or audit work. Run:
ls ~/.claude/projects/-Users-biznomad/memory/ | \
grep -iE "^(feedback|playbook|finding|project)_(hv|holistic|vitalis|vicelle|dlluxe|liquid.wizdom|biznomad)_"
Read every match in full. These memories override defaults — copy
conventions, banned phrases, known-broken assets, attribution quirks,
approval rules. The Apr/May 2026 HV run violated 3 such memories
(mineral copy, pixel walkthrough, attribution framing) by not doing
this step. See [[feedback_triple_check_against_memory]].
Also read the always-on rules: [[feedback_meta_ads_72h_blackout]],
[[feedback_approve_before_posting]], [[feedback_no_live_push]],
[[feedback_image_quality]], [[feedback_product_reference_images]].
-
Apply 72h blackout. If any account change happened in the last 72h,
only run read-only phases (1, 3, 5). Phases 2 and 4 produce local
assets — they do not push to Meta — so they are always safe. Phase 5
is read-only audit.
-
Verify infrastructure state with live checks, not docs. Before
citing "the token is 403'd" / "the pixel is firing" / "CAPI is down,"
SSH and curl/check yourself. Inherited diagnoses are often stale. See
[[finding_hv_overattribution_two_problems]] for the canonical example.
The 5 phases
Run sequentially. Each phase's output feeds the next. Stop and surface
findings to the user between phases — do NOT auto-launch ads.
Phase 1 — SPY (competitor active ads)
Invoke /ads-competitor scoped to Meta only. Pull every active ad from 3–5
named competitors with hooks, CTAs, offers, format, and "days running"
(a proxy for what's working). Save the swipe file to
./meta-ads-team/01-spy/swipe-file.md.
Inputs needed:
- 3–5 competitor brand names or Facebook page URLs
- Geo (default: US)
Output: ranked swipe file with hook patterns, offer patterns, and any
long-running ads (>30 days = likely winners).
Phase 2 — BULK CREATIVE (20 copy variations)
Invoke /ads-create for campaign concepts, then /ads-generate for the
actual copy. Requires brand-profile.json in the client root — if missing,
run /ads-dna first to build it.
Target: 20 on-brand copy variations across primary text + headlines, mapped
to messaging pillars from the brand DNA. Save to
./meta-ads-team/02-creative/copy-deck.md.
If visual assets are also needed, follow with /ads-photoshoot or
/higgsfield-product-photoshoot (preferred for HV / Vicelle / D'Lluxe
products — see [[reference_higgsfield_mcp]] and the relevant client product
visual specs, e.g. [[project_hv_product_visual_specs]]).
Phase 3 — ADSCORES (pre-launch quality gate)
Score every creative from Phase 2 BEFORE any spend. Invoke /ads-creative
with the Phase 2 outputs as input. The skill grades each ad on:
- Hook strength (first 3 seconds / first line)
- Copy clarity and CTA
- Emotional resonance vs. brand DNA
- Offer specificity
- Visual / copy fit
- Platform-native compliance (1:1, 4:5, 9:16 safe zones)
Output: ./meta-ads-team/03-scores/scorecard.md with PASS / WARN / FAIL per
creative. Only PASS-tier creatives proceed to launch.
Phase 4 — COMPETITOR ANGLE EXTRACTION
Re-run /ads-competitor in "angle extraction" mode against the Phase 1 swipe
file. Outputs: ranked hook categories (pain, status, curiosity, social
proof, mechanism), offer categories, and a gap list of angles competitors
are NOT running.
Use the gap list to brief one additional creative batch — these are the
highest-leverage tests because nobody else has tried them.
Save to ./meta-ads-team/04-angles/gap-analysis.md.
Phase 5 — ADS-META FULL AUDIT (186-check health score)
Invoke /ads-meta on the live account. Runs the full audit covering:
- Pixel + CAPI health, EMQ scores
- Creative fatigue and frequency
- Audience overlap and saturation
- Account structure (Advantage+ vs. manual)
- Bidding strategy and learning phase health
Output: ./meta-ads-team/05-audit/health-score.md with overall score 0–100
and prioritized fix list (P1 revenue / compliance / data, P2 efficiency,
P3 polish).
For HV specifically, cross-reference [[project_hv_audit_2026-04-26]] for
known issues (Klaviyo tracking gap, declining ROAS context).
Deliverable
After all 5 phases, produce one consolidated brief at
./meta-ads-team/SUMMARY.md with:
- Top 3 competitor patterns worth stealing (from Phase 1)
- Top 5 scored creatives ready to launch (from Phase 3)
- Top 3 untapped angles to test (from Phase 4)
- Top 3 P1 fixes on the live account (from Phase 5)
- Recommended 7-day test budget split (winners + gap angles)
Surface this to the user and STOP. Do not push to Meta. Per
[[feedback_approve_before_posting]] and [[feedback_meta_ads_72h_blackout]],
every launch and budget change requires explicit approval.
Client-specific notes
- HV (
Clients/Holistic-Vitalis/) — pixel/CAPI was broken Apr 21; verify
in Phase 5 before trusting any historical metrics.
- Vicelle (
Clients/Vicelle-Naturals/) — $25/day Mushroom Gummies ASC
baseline; DM ecosystem should improve ROAS from 1.29x to 2–3x.
- D'Lluxe (
Clients/DLluxe-Scrubs/) — new account, no historical data;
Phase 5 audit will mostly find learning-phase issues.
- Biznomad internal (
Clients/Biznomad/) — dental AGaaS pilot live at
$300/day CBO ([[project_dental_ads_launch]]); high-stakes account.
