Skip to content

fsbtactic-code/marketing-brain-skill / skills/marketing-brain

marketing-brain

Flagged

Create, improve, research, or audit marketing strategy and execution using a bundled marketing library, verified live primary sources for volatile facts, and clearly labeled agent judgment. Use for full marketing systems, audits, landing pages and CRO, competitor intelligence, brand voice, audience and JTBD research, email sequences, SMM and social content systems, paid ads, headlines, copywriting, positioning, offers, launches, GTM, growth, funnels, retention, pricing, SEO, analytics, attribution, experiments, and consumer or buyer psychology. Trigger for English and Russian requests, including requests for ready-to-use assets.

FileKindSize
SKILL.mdskill md10.6 KB
agents/openai.yamlconfig325 B
assets/BUILD-MANIFEST.jsonconfig1.2 KB
assets/cards.jsonlother21.1 KB
assets/edges.jsonlother33.0 KB
assets/knowledge.sqlitebinary136.0 KB
assets/legacy-library.mbpack.parts/manifest.jsonconfig857 B
assets/legacy-library.mbpack.parts/part-0000.binbinary48.0 MB
assets/legacy-library.mbpack.parts/part-0001.binbinary48.0 MB
assets/legacy-library.mbpack.parts/part-0002.binbinary7.7 MB
assets/research-catalog.jsonlother280.4 KB
assets/source-channels.jsonlother52.1 KB
assets/sources.jsonlother10.2 KB
assets/templates/content-calendar.mdmarkdown1.3 KB
assets/templates/evidence-ledger.mdmarkdown1.3 KB
assets/templates/experiment-backlog.mdmarkdown1.2 KB
assets/templates/issue-registry.mdmarkdown1.2 KB
assets/templates/marketing-brief.mdmarkdown1.6 KB
assets/templates/marketing-system-manifest.jsonconfig1.1 KB
assets/wiki/cards/card_ad_effects_heterogeneous_by_firm.mdmarkdown1.3 KB
assets/wiki/cards/card_ad_roi_measurement_noise.mdmarkdown1.4 KB
assets/wiki/cards/card_choice_overload_context_dependent.mdmarkdown1.4 KB
assets/wiki/cards/card_cost_transparency_context.mdmarkdown1.2 KB
assets/wiki/cards/card_ftc_material_connection_disclosure.mdmarkdown1.4 KB
assets/wiki/cards/card_google_ads_experiment_hygiene.mdmarkdown1.3 KB
assets/wiki/cards/card_google_no_word_count_freshness_trick.mdmarkdown1.2 KB
assets/wiki/cards/card_google_people_first.mdmarkdown1.2 KB
assets/wiki/cards/card_google_who_how_why.mdmarkdown1.1 KB
assets/wiki/cards/card_incrementality_control_group.mdmarkdown1.2 KB
assets/wiki/cards/card_incrementality_precision_stratification.mdmarkdown1.3 KB
assets/wiki/cards/card_short_video_ad_spillover.mdmarkdown1.3 KB
assets/wiki/cards/card_telegram_ad_format.mdmarkdown1.0 KB
assets/wiki/cards/card_telegram_destination_quality.mdmarkdown1.1 KB
assets/wiki/cards/card_telegram_no_unsupported_absolutes.mdmarkdown1.0 KB
assets/wiki/cards/card_viral_content_arousal.mdmarkdown1.3 KB
assets/wiki/cards/card_youtube_audience_signals.mdmarkdown1.2 KB
assets/wiki/cards/card_youtube_publish_time.mdmarkdown1.4 KB
assets/wiki/cards/card_youtube_quality_over_frequency.mdmarkdown1.3 KB
assets/wiki/cards/card_youtube_search_ranking.mdmarkdown1.1 KB
assets/wiki/cards/card_youtube_tags_low_priority.mdmarkdown1019 B
assets/wiki/domains/consumer-behavior.mdmarkdown2.2 KB
assets/wiki/domains/content-editorial.mdmarkdown5.1 KB
assets/wiki/domains/creator-community.mdmarkdown1009 B
assets/wiki/domains/cro-funnels.mdmarkdown2.3 KB
assets/wiki/domains/law-policy-ethics.mdmarkdown2.8 KB
assets/wiki/domains/measurement-analytics.mdmarkdown4.5 KB
assets/wiki/domains/offer-messaging-copy.mdmarkdown2.7 KB
assets/wiki/domains/organic-smm.mdmarkdown4.9 KB
assets/wiki/domains/paid-acquisition.mdmarkdown6.2 KB
assets/wiki/domains/pricing-promotions.mdmarkdown850 B
assets/wiki/domains/seo-geo-ai-search.mdmarkdown2.8 KB
assets/wiki/domains/strategy-research.mdmarkdown932 B
assets/wiki/index.mdmarkdown935 B
assets/wiki/open-questions.mdmarkdown159 B
assets/wiki/pages.jsonlother18.3 KB
references/research-catalog-policy.mdmarkdown1.3 KB
references/retrieval-policy.mdmarkdown1.0 KB
references/task-playbooks.mdmarkdown5.1 KB
references/taxonomy.mdmarkdown1005 B
references/third-party-notices.mdmarkdown1.6 KB
references/workflow-audience.mdmarkdown8.4 KB
references/workflow-audit.mdmarkdown7.0 KB
references/workflow-brand-voice.mdmarkdown8.0 KB
references/workflow-competitors.mdmarkdown7.2 KB
references/workflow-email.mdmarkdown8.8 KB
references/workflow-full-system.mdmarkdown9.2 KB
references/workflow-headlines.mdmarkdown8.4 KB
references/workflow-landing-cro.mdmarkdown6.9 KB
references/workflow-paid-ads.mdmarkdown10.1 KB
references/workflow-social.mdmarkdown9.2 KB
scripts/marketing_context_output.pyscript7.3 KB
scripts/marketing_context_query.pyscript8.6 KB
scripts/materialize_legacy_pack.pyscript12.1 KB
scripts/mbpack_codec.pyscript4.1 KB
scripts/mbpack_output.pyscript5.2 KB
scripts/retrieve_marketing_context.pyscript13.1 KB
scripts/search_knowledge.pyscript7.2 KB
scripts/search_legacy_pack.pyscript131 B
scripts/search_private_pack.pyscript11.0 KB
scripts/search_research_catalog.pyscript6.2 KB
scripts/verify_public_install.pyscript7.1 KB

references/workflow-headlines.md

View on GitHub
# Headline and Copy Testing Workflow

Use this workflow for headlines, hooks, subject lines, ad titles, landing page hero copy, social openers, video hooks, and copy variants intended for testing.

## Trigger conditions

Run this workflow when the user asks to:

- generate or improve headlines for a named surface;
- create testable hook or subject-line variants;
- adapt one message to several awareness states or channels;
- select the strongest headline hypotheses;
- diagnose why current headlines are unclear or unconvincing;
- create a copy test plan rather than an unranked idea dump.

For a single mechanical edit, preserve the user's meaning and do not expand into a full lab unless requested.

## Status model

- `PASS`: the variant is supported, compliant, surface-ready, and testable.
- `NEEDS_DATA`: the variant is usable only after the named claim, proof, limit, or audience input is supplied.
- `BLOCKED`: the variant must not be published or tested in its current form.

Do not create a fictional conversion score.

## Required inputs

- Target surface and platform.
- Primary objective and desired action.
- Audience and awareness state.
- Product, offer, and relevant customer situation.
- Differentiated mechanism or reason to choose.
- Available proof.
- Brand voice and prohibited language.
- Claims that require substantiation.

## Optional inputs

- Current headline and performance baseline.
- Traffic source and message shown before the headline.
- Character or layout constraints.
- Competitor headlines for differentiation analysis.
- Existing `BRAND-VOICE.md`, `AUDIENCE-RESEARCH.md`, `POSITIONING-AND-OFFER.md`, or `COMPETITOR-REPORT.md`.
- Past tests, sample size, conversion event, and guardrail metrics.
- Localization and accessibility requirements.

## Dependencies and precedence
Approved claims and proof outrank cleverness. Message match outranks novelty. Brand voice constrains expression but does not justify unsupported promises. Platform limits must be checked before finalizing copy.

Do not generate an arbitrary quota of 100 headlines. Produce the smallest diverse set needed for the decision, normally 12 to 30, and do not pad the set with superficial rewrites.

## Retrieval plan

Run one to three focused queries from the directory containing `SKILL.md`:

```text
python scripts/retrieve_marketing_context.py --query "headline clarity message hierarchy offer proof and copy testing" --domain offer-messaging-copy --limit 5 --json
python scripts/retrieve_marketing_context.py --query "attention comprehension choice and buyer awareness" --domain consumer-behavior --limit 5 --json
python scripts/retrieve_marketing_context.py --query "ad creative hooks platform message match and measurement" --domain paid-acquisition --limit 5 --json
```

Choose queries based on the surface. Do not treat a swipe file, template, or creator example as causal evidence.

## Live checks

Verify current character limits, truncation behavior, restricted claims, advertising policies, and formatting rules using official platform documentation. Record the source URL and check date. Treat algorithm advice without an official source as a hypothesis.

## Workflow phases

### Phase 1: Define the decision

State where the headline appears, what the visitor already knows, the single next action, and how success will be measured. Identify the preceding message and the content immediately after the headline.

### Phase 2: Build the claim registry

List every factual, comparative, numerical, testimonial, deadline, scarcity, guarantee, and authority claim that could appear. Mark each as approved, needs proof, prohibited, or not applicable. Replace missing proof with a neutral formulation or a labeled placeholder that names the missing input and owner.

### Phase 3: Create the message spine

Define:

- audience situation;
- desired progress;
- primary value;
- differentiated mechanism;
- strongest proof;
- main objection;
- desired action;
- tone boundary.

If these cannot be stated without guessing, request the missing input or mark the related output `NEEDS_DATA`.

### Phase 4: Build an angle matrix

Select only angles relevant to the awareness state and surface. Possible angles include outcome, problem recognition, mechanism, proof, comparison, objection, demonstration, category reframing, risk reduction, and direct offer.

For each angle state the audience belief it tests and the evidence it can honestly use. Do not require urgency, fear, controversy, curiosity, or a cognitive trigger in every headline.

### Phase 5: Draft strategically distinct variants

Create several variants per selected angle. Vary the underlying proposition, proof, specificity, or framing. Do not count punctuation, synonym swaps, or word order changes as separate hypotheses.

Keep labeled placeholders visible for unverified figures, customer names, awards, deadlines, or guarantees only in `NEEDS_DATA` or `BLOCKED` output. Never invent these values to make the copy sound finished. A `PASS` result cannot contain unresolved placeholders.

### Phase 6: Apply production filters

Check each variant for clarity, message match, audience fit, proof, voice, grammar, specificity, accessibility, policy compliance, and actual character count. Remove clickbait that withholds the promised value or creates a misleading implication.

### Phase 7: Select finalists

Choose three to five finalists that test meaningfully different beliefs. Explain why each is worth testing and what result would teach the team. Do not predict a winner or attach an invented conversion probability.

### Phase 8: Design the test

Define baseline, primary metric, guardrails, unit of randomization, minimum detectable effect, sample-size method, run length, conversion delay, stopping rule, and interpretation. If traffic data is missing, provide the formula and inputs needed instead of a sample-size guess.

## Exact output contract

When file output is requested or artifact mode is active, save the final artifact as `HEADLINE-LAB.md`. Otherwise return the same contract in the response. Use these sections in this order:

1. `Decision, surface, version, and overall status`
2. `Inputs, assumptions, and source coverage`
3. `Live platform checks`
4. `Claim registry`
5. `Message spine`
6. `Angle matrix`
7. `Candidate variants grouped by strategic hypothesis`
8. `Rejected patterns and reasons`
9. `Finalists with rationale and status`
10. `Surface-ready copy with verified character counts`
11. `Test design and decision rules`
12. `Quality gate results`
13. `Missing data and next owner action`

Each candidate row must include variant ID, copy, angle, awareness state, proof used, character count, status, and intended learning.

## Quality gates

- Message match: the headline fits the preceding source and following content.
- Strategic diversity: variants test different beliefs rather than cosmetic wording.
- Claim integrity: every factual or numerical claim is approved and traceable.
- No fake proof: no invented users, results, studies, awards, quotes, or rankings.
- No fake urgency: scarcity and deadlines appear only when operationally true.
- Surface fit: character limits and truncation behavior were checked live.
- Voice fit: language follows the approved guide without becoming generic.
- Clarity: a target reader can identify the offer or relevant problem quickly.
- Test validity: the plan has a metric, guardrail, decision rule, and sample-size method.
- No fake winner: finalists are hypotheses, not predicted performance claims.

## Stop and fallback rules

- If the surface is unknown, ask where the copy will appear before final formatting.
- If the offer or audience is undefined, provide a short discovery brief and mark `BLOCKED`.
- If proof for a material claim is missing, neutralize the claim or use a labeled placeholder with the required owner and mark `NEEDS_DATA`.
- If platform rules cannot be verified, deliver drafts but mark production readiness `NEEDS_DATA`.
- If the requested tactic requires deception, fake scarcity, fabricated authority, or unsupported results, refuse that tactic and offer a truthful angle.
- If baseline and traffic are missing, do not provide a fixed sample size or guaranteed test duration.

## Completion criteria
The workflow is complete when `HEADLINE-LAB.md` exists or the complete contract is returned in the response, all variants are meaningfully distinct and traceable, platform constraints are checked or marked, every gate has a status, no unsupported claims remain, and the team has a valid next test or a precise data request.
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 194kUpdated Claude CodeCodexVerifiedInstall