openclaw-frontier-stack

Use when an ecomm operator wants to design or evaluate a controlled A/B test. Triggers when the user mentions "A/B test", "split test", "is this test significant", "sample size", "sequential testing", "MDE", "minimum detectable effect", or "should I stop the test". For CRO test ideas, see cro. For ad creative tests specifically, see ad-creative. For email-marketing test design, see email-marketing.

Use when an ecomm operator wants to audit WCAG 2.2 AA conformance, evaluate ADA litigation risk, choose an accessibility audit vendor, set up screen reader testing, or build the business case for accessibility investment. Triggers when the user mentions "accessibility", "WCAG", "ADA", "ADA compliance", "screen reader", "accessibility audit", "a11y", "Section 508", "AODA", "EAA", "accessibility lawsuit", "demand letter", or "accessibility ROI". For broader CRO and conversion-rate work that accessibility intersects with, see cro. For SEO outcomes that accessibility supports, see seo-audit. For broader change-management when accessibility becomes a cross-team workstream, see change-management-policy.

Use when an ecomm operator wants to design or iterate ad creative — images, video, carousels, UGC. Triggers when the user mentions "ad creative", "creative for Meta ads", "UGC ads", "what should the hook be", "creative testing", "hook-body-cta", "first 3 seconds", or "our creative is fatigued". For channel-strategy decisions, see ads. For copy in the creative, see copywriting. For controlled test methodology, see ab-testing.

Use when an ecomm operator wants paid-acquisition strategy across channels — Meta, Google, TikTok, YouTube, native — including budget allocation, channel mix, scaling, and pause decisions. Triggers when the user mentions "ad strategy", "channel mix", "should I run Google or Meta", "scaling paid", "budget allocation", "diminishing returns", or "should I pause this channel". For tactical creative design, see ad-creative. For LTV-driven channel evaluation, see paid-ltv-optimization. For controlled-test methodology, see ab-testing.

Use when an ecomm operator wants to design, audit, or relaunch an affiliate program — in-house affiliate management, affiliate network (Impact, Rakuten, Awin, CJ, ShareASale), publisher partnerships, content-affiliate, coupon-affiliate. Triggers when the user mentions "affiliate program", "affiliate marketing", "publisher partnerships", "Rakuten", "Impact", "Awin", "affiliate network selection", "publisher recruitment", or "is our affiliate program profitable". For advocate-driven referrals from existing customers, see referral-program-design. For sponsored creator content, see influencer-program-design. For broader social-channel strategy, see social-strategy.

Use when an ecomm operator wants to generate product images, marketing creative, lifestyle scenes, or static social assets using AI image-generation models — choosing between models, designing prompts, maintaining brand consistency, navigating disclosure and copyright. Triggers when the user mentions "generate product image", "AI image", "marketing image", "creative variants", "model selection for image", "Gemini image", "Flux", "Ideogram", "ChatGPT Images", "Midjourney", "Recraft", or "Stable Diffusion". For the in-feed creative strategy that uses the generated image, see ad-creative. For the headline and CTA copy in the image, see copywriting. For video generation, see ai-video-generation.

Use when an ecomm operator wants visibility in AI-driven search surfaces — ChatGPT, Claude, Perplexity, Google AI Overviews — not just classic Google blue links. Triggers when the user mentions "AI SEO", "LLM SEO", "ChatGPT search", "AI Overviews", "llms.txt", "agentic search", "query fan-out", or "I want my brand to show up when people ask AI for product recommendations". For classic technical and on-page SEO, see seo-audit. For mass page generation, see programmatic-seo. For structured-data design, see schema-markup.

Use when an ecomm operator wants to generate marketing video using AI models — product hero shots, social-first 9:16, animated explainers, lifestyle B-roll, talking-head video. Triggers when the user mentions "generate video", "AI video", "Sora", "Veo", "Runway", "Kling", "short-form video", "product video generation", "AI video for ads", or "video creative testing". For the in-feed creative strategy that uses the generated video, see ad-creative. For AI image generation as a sibling workflow, see ai-image-generation. For social-channel-native video strategy, see social-strategy.

Use when retiring an API surface — endpoint, field, header, parameter, or behavior — that has existing consumers. Triggers when the user mentions "deprecate this endpoint", "sunset this field", "remove this API", "v1 retirement", "breaking change", "deprecation timeline", "communication plan", or "they hadn't migrated". The skill covers the deprecation lifecycle (announce, mark, instrument, migrate, remove), sunset windows for internal vs public consumers, RFC 8594 Sunset/Deprecation headers, usage telemetry, and recovery when consumers don't move. For the design of the API that's now being deprecated, see api-design. For the broader release-gate question of safety in shipping these changes, see safe-public-release.

Use when designing a new HTTP/RPC API or evolving an existing one without breaking consumers. Triggers when the user mentions "design this API", "REST vs GraphQL", "endpoint design", "API versioning", "breaking change", "API contract", "should this be one endpoint or two", "pagination", or "error response shape". The output is a contract that consumers can build against confidently and that the team can evolve without scheduled flag days. For the durable record of the resulting decisions, see architecture-decision-records. For the storage layer those endpoints touch, see schema-design.

Use when capturing a non-trivial architecture decision so future readers know why the system is the way it is. Triggers when the user mentions "write an ADR", "architecture decision record", "should I document this decision", "we're choosing between X and Y", "design decision", or asks "why did we do it this way". The bar is: would a future reader be confused without this record. For the deeper design work that produces decisions, see api-design and schema-design. For sharing the decision through review, use requesting-code-review.

Use when an ecomm operator wants to choose, evaluate, or replace an attribution model — first-touch vs last-touch vs multi-touch, incrementality testing, MMM, MTA, post-iOS-14 measurement. Triggers when the user mentions "attribution model", "first-touch vs last-touch", "multi-touch attribution", "incrementality testing", "MMM", "media mix model", "MTA", "how do I trust my attribution", or "our attribution is broken". For paid-channel decisions downstream of attribution, see paid-ltv-optimization. For tactical channel-level guidance, see ads. For cohort-level retention analysis, see cohort-retention.

Use when designing, auditing, or testing a backup-and-restore system — choosing RPO/RTO targets, scheduling drills, verifying restores, or recovering from disaster. Triggers when the user mentions "backup strategy", "restore drill", "RPO", "RTO", "point-in-time recovery", "disaster recovery", "did the backup actually work", "we lost the database", "retention policy", or "encryption key". The skill covers RPO/RTO sizing, restore-verification cadence, dependency ordering on recovery, key recovery for encrypted backups, and the principle that a backup you haven't restored is not a backup. For the incident response when production is down, see incident-response. For the data-corruption playbook inside an incident, see incident-response's data-corruption section.

Use when an ecomm operator wants to design or optimize bundles, raise AOV, or build good-better-best tiering. Triggers when the user mentions "bundle pricing", "should I bundle", "AOV is low", "increase order value", "kit pricing", "good better best", or "what should I sell together". For repeat-purchase logic at the second order, see repeat-purchase. For paid acquisition economics tied to bundle AOV, see paid-ltv-optimization.

Use when sizing a service for the next quarter of demand growth, deciding how much headroom to hold, deciding whether to scale before or during traffic, or recovering from an incident where the service ran out of capacity in production. Triggers when the user mentions "capacity planning", "how many replicas", "what size machine", "headroom", "scaling headroom", "size for peak", "burst capacity", "sustained capacity", "Black Friday sizing", "we ran out of capacity", "the load test didn't predict this", or "the autoscaler couldn't catch up". The skill covers demand forecasting, the difference between burst and sustained capacity, headroom rules under uncertainty, load-test inputs that match production, and the failure modes when capacity planning is treated as a one-time exercise. For the load-testing mechanics, see load-testing. For the alerts that detect capacity exhaustion, see monitoring-and-alerting. For the SLO impact of capacity shortfalls, see slo-design.

Use when an ecomm operator wants to reduce cart abandonment rate, recover lost checkout revenue, or design abandonment flows. Triggers when the user mentions "cart abandonment", "people aren't completing checkout", "abandonment flow", "exit intent", "cart recovery email", "checkout drop-off", or "my abandonment rate is too high". For repeat-purchase flow design (post first purchase), see repeat-purchase. For paid retargeting LTV math, see paid-ltv-optimization.

Use when defining or refining the policy that gates production changes — who can ship what, when review is required, what evidence the change must carry, and how the review actually happens. Triggers when the user mentions "change advisory board", "CAB", "change ticket", "change policy", "who needs to approve this", "change risk tiers", "do we need a freeze for this", "rubber-stamp approvals", or "we keep approving changes that break production". The skill covers risk-tiered approval rules, the evidence each tier requires, CAB anti-patterns (theater vs real review), and the emergency-change carve-out. For the deploy mechanics once a change is approved, see safe-public-release. For the post-incident review when an approved change broke production, see post-mortem-writing.

Use when you are the reviewer on someone else's pull request — deciding what to comment on, how to phrase the comment, and whether to approve, request changes, or leave it open. Triggers when the user mentions "review this PR", "give code review", "what should I comment on", "is this ready to merge", "approve with comments", "should I request changes", or "drive-by review". The job is to protect the codebase without burying the author. For the author's side of the same conversation, see requesting-code-review. For the broader question of whether the change is safe to ship at all, see safe-public-release.

Use when you are the author of a pull request reading reviewer feedback — deciding what to address, what to push back on, when to ask for clarification, and how to close the loop. Triggers when the user mentions "responding to code review", "address PR feedback", "argue back on a review comment", "request reviewer change", "I disagree with this review", "how do I respond to this nit", or "this review is wrong". The goal is to land good changes without burning the reviewer, and to disagree productively when the reviewer is wrong. For the reviewer's side of the same conversation, see code-review-giving. For the prep that makes the PR easy to review in the first place, see requesting-code-review.

Use when an ecomm operator wants to analyze retention curves by acquisition cohort, diagnose whether cohorts are getting worse over time, or compare retention by acquisition channel. Triggers when the user mentions "cohort analysis", "retention curves", "retention by channel", "cohort survival", "are my cohorts getting worse", "first-touch retention", or "cohort quality". For paid acquisition channel decisions, see paid-ltv-optimization. For subscription churn diagnosis, see subscription-churn. For lapsed customer reactivation, see winback-flows.

Use when an ecomm operator wants to design a brand community, set tier benefits, build a UGC engine, choose between Discord/Slack/Circle/Facebook, or audit the metrics they track on an existing community. Triggers when the user mentions "brand community", "community program", "VIP community", "Discord community", "Slack community", "Circle community", "Facebook group", "UGC program", "ambassador program", or "what's a good community KPI". For paid creator partnerships outside community, see influencer-program-design. For affiliate-link-driven communities, see affiliate-program-design. For points-based loyalty programs not community-driven, see loyalty-program-design.

Use when an ecomm operator wants to design content production across SEO, social, email, and ads. Triggers when the user mentions "content strategy", "editorial calendar", "blog plan", "content marketing", "topic clusters", "pillar pages", or "we need to publish more". For AI-search content design, see ai-seo. For email cadence design, see email-marketing. For organic social specifically, see social-strategy.

Use when an ecomm operator wants to write or rewrite product copy, headlines, CTAs, ad copy, email subject lines, or descriptions. Triggers when the user mentions "write copy", "rewrite this", "headline", "CTA", "subject line", "product description", "tagline", or "this copy is flat". For PDP conversion as a whole, see cro. For ad-specific creative pairing, see ad-creative. For email body design beyond subject lines, see email-marketing.

Use when an ecomm operator is choosing a CRM, building a single-customer view, designing lead scoring, defining lifecycle stages, or fixing a marketing-to-sales handoff. Triggers when the user mentions "CRM", "HubSpot", "Salesforce", "single customer view", "lead scoring", "lifecycle stages", "MQL", "SQL", "marketing sales handoff", "customer 360", or "we need a CRM". For the broader CDP question of identity resolution and event collection, see customer-data-platform-strategy. For the support and Ticketing side of CRM, see post-purchase-experience. For the email and SMS execution layer, see email-marketing.

Use when an ecomm operator wants to lift conversion rate on a PDP, cart, checkout, or signup. Triggers when the user mentions "improve conversions", "PDP conversion rate", "cart drop-off", "checkout optimization", "this page isn't converting", or "CRO". For abandonment recovery flows after a customer leaves, see cart-abandonment-recovery. For LTV-driven channel decisions upstream, see paid-ltv-optimization. For controlled-test methodology, see ab-testing.

Use when an ecomm operator wants to design or audit cross-sell sequences using SKU affinity data, raise attach rate, or pick the right post-purchase add-on to surface. Triggers when the user mentions "what should I cross-sell with X", "cross-sell flow", "SKUs frequently bought together", "post-purchase cross-sell", "increase attach rate", "in-cart upsell", or "what to recommend after checkout". For multi-SKU choice architecture and good-better-best, see bundle-pricing. For second-order replenishment timing, see repeat-purchase. For abandoned-cart recovery rather than completed-purchase add-on, see cart-abandonment-recovery.

Use when an ecomm operator is evaluating whether to buy a customer data platform (CDP), build one, or stay on a warehouse + reverse-ETL stack. Triggers when the user mentions "CDP", "customer data platform", "Segment vs Rudderstack", "buy vs build CDP", "identity resolution", "single customer view", "should we get a CDP", "event schema", "data activation", "compliance retention", or "warehouse-first stack". For the survey infrastructure that often gets bundled into a CDP decision, see customer-research. For the downstream paid-channel activation that depends on CDP audience export, see paid-ltv-optimization. For email and SMS activation surfaces specifically, see email-marketing.

Use when an ecomm operator wants qualitative research on customers — interviews, JTBD, voice-of-customer mining. Triggers when the user mentions "customer research", "interviews", "JTBD", "Jobs to Be Done", "voice of customer", "why do customers buy", "user research", or "we need to understand our customer better". For positioning that consumes the research, see product-marketing-positioning. For copy that uses customer language, see copywriting. For conversion implications, see cro.

Use when assigning a data tier to a new dataset, deciding what storage and access controls a field needs, or auditing whether the team's current handling of sensitive data meets the policy. Triggers when the user mentions "data classification", "PII", "personally identifiable information", "data tiers", "public internal confidential restricted", "data retention", "data handling policy", "right to be forgotten", "data access controls", "audit log for data access", "what data can we store", "encryption at rest", "field-level encryption", or "we have customer data in a debug log". The skill covers the four-tier classification scheme, PII handling specifics, retention rules per tier, access controls including audit logging, and the anti-patterns that arise when classification is treated as compliance theater. For threat modeling around data exposure, see threat-modeling. For secrets storage, see secrets-management. For the security review of changes that touch sensitive data, see security-review.

Use when planning or executing a schema migration on a production database — adding columns, changing types, dropping fields, adding indexes, or shifting constraints without downtime or data loss. Triggers when the user mentions "database migration", "schema change", "online migration", "expand contract", "blue-green schema", "Alembic", "Flyway", "Liquibase", "gh-ost", "pt-online-schema-change", or "this migration is locking the table". The skill covers the expand-contract pattern, lock-free techniques, the online-safety filter per operation type, backfill strategy, and reversible-vs-one-way migrations. For backup and restore in case the migration goes wrong, see backup-and-restore. For the schema design that precedes the migration, see schema-design.

Use when upgrading npm/cargo/pip/maven/Go module/etc. dependencies without breaking production. Triggers when the user mentions "upgrade dependencies", "bump version", "npm update", "Dependabot PR", "security patch", "renovate", "is this safe to merge", or "should I take this update". The output is a defensible upgrade plan that catches the actual risks rather than rubber-stamping. For the security side of risky deps, see security-review. For the refactoring that may be needed when a major version lands, see refactoring-safety.

Use when designing, scheduling, or running a disaster recovery exercise — tabletops, live drills, chaos engineering, GameDays. Triggers when the user mentions "DR drill", "disaster recovery exercise", "tabletop", "chaos engineering", "GameDay", "region failover drill", "we should test our DR plan", or "what would happen if we lost X". The skill covers tabletop vs live tradeoffs, failure modes to drill, blast radius framing, RTO/RPO validation, follow-up discipline, and the chaos-vs-DR distinction. For the underlying backup-restore system the drills validate, see backup-and-restore. For the incident response that DR exercises rehearse, see incident-response.

Use when an ecomm operator wants to design or audit a dunning sequence, recover failed payments on subscriptions or one-time orders, lower involuntary churn, tune retry schedule, or build pre-failure prevention. Triggers when the user mentions "failed payments", "dunning sequence", "card declined", "subscription payment failed", "involuntary churn", "payment retry strategy", or "expired card recovery". For voluntary subscription churn distinct from involuntary, see subscription-churn. For broader lapsed-customer reactivation that is downstream of dunning failure, see winback-flows.

Use when a task will span multiple subagents, multiple sessions, or multiple files. Append-only JSONL ledger of task claims, path claims, and result receipts prevents parallel agents from stomping on each other. Triggers when the user dispatches parallel subagents (often via superpowers' dispatching-parallel-agents skill), kicks off a multi-step refactor across files, or asks "who owns this task?" or "what's the state of X?". For in-session task tracking, prefer TodoWrite. For coordination across sessions or subagents, use this skill.

Use when an ecomm operator wants broader email strategy beyond specific flows — broadcasts, segmentation, list health, deliverability, the owned-channel discipline. Triggers when the user mentions "email strategy", "email marketing", "newsletter", "list growth", "deliverability", "inbox placement", "email segmentation", or "our email program is flat". For abandonment flows specifically, see cart-abandonment-recovery. For winback flows, see winback-flows. For subscription churn flows, see subscription-churn. For dunning, see dunning-deep-dive.

Use when designing an A/B test or experiment behind a feature flag — picking the primary metric, computing sample size and MDE, deciding frequentist vs Bayesian, planning analysis before the test runs. Triggers when the user mentions "experiment design", "A/B test statistics", "MDE", "minimum detectable effect", "sample size calculation", "experiment platform", "Bayesian vs frequentist", "CUPED", "guardrail metrics", "peeking", or "we ran an A/B test, what does it mean". The skill covers pre-registration discipline, MDE, peeking risks, variance reduction, guardrail metrics, and heterogeneous treatment effects. For the rollout mechanism (the flag itself), see feature-flagging. For the broader release workflow the experiment lives inside, see safe-public-release.

Use when adding, designing, or removing a feature flag. Triggers when the user mentions "feature flag", "rollout strategy", "kill switch", "gradual rollout", "percentage rollout", "cohort flag", "remove this flag", "flag debt", "sticky bucket", or "is this flag still needed". The skill covers flag taxonomy, default-state decisions, kill-switch design, percentage targeting, telemetry-on-flags, and the deletion lifecycle that prevents flag debt. For the broader release workflow flags exist inside, see safe-public-release. For the on-call response when a flag flip is part of mitigation, see incident-response.

Use when an ecomm operator is considering moving to headless commerce, evaluating a replatform, or weighing the team-shape requirements of a custom storefront. Triggers when the user mentions "headless", "headless commerce", "should we go headless", "Hydrogen vs Next", "custom storefront", "decoupled frontend", "replatform", "Shopify Plus headless", "BigCommerce headless", "Commerce Layer", "Saleor", "performance budget", or "we want full control of the frontend". For the SEO consequences of replatforming, see seo-audit. For the team and operational shape of running a custom stack, see local-dev-environment. For the migration mechanics of moving carts/customers/orders, see database-migration-safety.

Use when production is on fire and you need to triage, mitigate, and communicate. Triggers when the user mentions "production is down", "users are reporting errors", "we have an incident", "Sev 1", "Sev 2", "page from on-call", "alert just fired", "things are broken in prod", or "rollback now". The goal here is to stop the bleeding, preserve evidence, and run clean comms — not to find the root cause. For root cause work after mitigation, see root-cause-analysis. For writing the post-mortem afterward, see post-mortem-writing. For designing the alerts that paged you in the first place, use monitoring-and-alerting.

Use when an ecomm operator wants to design, audit, or scale an influencer or creator program — paid creators, UGC partners, ambassador tiers, gifted-only seeding. Triggers when the user mentions "influencer marketing", "creator program", "TikTok creators", "Instagram influencers", "UGC creators", "macro vs micro influencer", "creator brief", "should I pay creators", or "is our influencer spend working". For the in-feed creative the creator produces, see ad-creative. For paid-channel scaling math around creator usage rights, see ads. For advocate-driven mechanics where existing customers refer friends, see referral-program-design.