agile_agentic_analytics

Use when the user wants to measure the impact of a campaign that promotes a NEW product to existing customers — phrases like "cross-sell analysis", "cross-sell campaign results", "product attach rate", "account open rate", "new product adoption from campaign", "second product campaign", "add-a-product campaign", "checking-to-savings campaign", "credit card cross-sell", "did the campaign drive new account opens", "attach campaign", or any review of a campaign whose primary goal is to get current customers to take on a product they don't already have. Also trigger on "what was our conversion rate on the X campaign to existing customers" and "compare treated vs holdout for new account opens". Cross-sell differs from up-sell: the headline outcome is a binary CONVERSION (did they open the new product?), and value lift is the revenue or balance from the new product, not growth on existing accounts. Handles both holdout-controlled designs (statistical testing on conversion rate AND value) and uncontrolled designs (d

Use when the user wants to measure the impact of a marketing campaign on existing clients, account owners, or current customers — phrases like "up-sell analysis", "did the campaign move balances", "expansion campaign results", "wallet share lift", "AUM lift", "deposit growth from campaign", "existing customer campaign", "customer growth campaign", "upgrade campaign", "in-book campaign", or any review of a campaign targeted at people who are already customers (so there is no acquisition or conversion-rate question). Also trigger on "how much did balances/sales/revenue go up for treated clients" and "compare treated vs holdout for existing customers". Handles both holdout-controlled designs (statistical testing on the difference) and uncontrolled pre/post designs (descriptive lift only, clearly flagged as non-causal).

Use when initializing, maintaining, validating, searching, documenting, or assembling context from a file-based knowledge base with the bundled vaultli CLI. Trigger on knowledge vaults, KB setup, .kbroot, YAML frontmatter, sidecar markdown, INDEX.jsonl, documenting SQL/templates/configs/scripts, validating stale or broken KB state, searching vault metadata, hydrating source assets, assembling retrieval context, or federated lookup across multiple vault roots.

Manage a file-based knowledge vault with YAML frontmatter, sidecar markdown, and JSONL indexing. Use when initializing or maintaining a knowledge base, attaching metadata to markdown or non-markdown assets, rebuilding or validating INDEX.jsonl, or searching vault records by metadata. Trigger on: "knowledge base", "vault", "frontmatter", "sidecar", "INDEX.jsonl", "document this query/template/skill", "search the vault", or "set up KB docs".

Use this skill when the user wants to plan an analysis before running it, scope analytical work, design an investigation, write an analysis plan, decide what data or metrics are needed, or turn a vague metric/business question into an executable analyst workplan. Trigger on phrases like "analysis planning", "plan this analysis", "how should we analyze", "what data do we need", "design the analysis", "scope this analysis", "analysis plan", "investigation plan", "before we pull data", or "senior analyst, help me plan this".

Use when the user wants to extract data, pull data from a source, load a CSV or JSON export into the workspace, connect a data source, refresh data, ingest raw platform exports, or normalize raw data for the marketing-analytics skills. Trigger on 'extract data', 'pull data from <platform>', 'load CSV/JSON into workspace', 'connect a data source', 'refresh data', 'normalize raw data', 'import my ad export', 'get the campaign data in', or 'set up the workspace'. This is the upstream skill every other marketing-analytics skill depends on: it lands source exports in workspace/raw/ and writes validated, normalized files to workspace/processed/. Run this first when a downstream skill reports a missing or malformed workspace input. When the user wants analysis, modeling, or reporting rather than getting data in, route to the matching analytics skill instead (see references/skill-index.md).

Owns email performance, engagement, and deliverability analytics. Use when the user mentions email performance, open rate, click rate, click-to-delivered rate, deliverability, bounce rate, unsubscribe rate, email engagement, lifecycle email, drip campaign, email automation, send-time optimization, subject line performance, SPF, DKIM, DMARC, email blocklist, inbox placement, or list health. Also trigger on 'how are our emails performing' or 'improve email engagement.' When NOT to use: for measuring whether email CAUSED a purchase (holdout / incrementality / lift), defer to the experimentation plugin's email-incrementality skill; for the statistics of any email A/B test, delegate to the experimentation skill. If send data is not yet in the workspace, run data-extraction first.

Use when the user mentions funnel, conversion funnel, drop-off, drop off, conversion rate, conversion optimization, CRO, bottleneck, funnel analysis, checkout flow, signup flow, onboarding funnel, activation funnel, abandonment, cart abandonment, form abandonment, user flow, step completion, or funnel comparison. Also trigger on 'where are we losing people' or 'why is conversion low.' For site traffic sources and session behavior use web-analytics; for running the A/B test that validates a CRO fix use the experimentation skill; this skill measures step-by-step drop-off and prioritizes bottlenecks. If event data is not yet in the workspace, run data-extraction first.

Top-level workflow for maintaining plugins in this marketplace. Use when creating a plugin, adding or moving plugin skills, checking plugin manifests, updating marketplace entries, validating packaging, preparing a plugin release, or deciding whether work belongs in a plugin, a plugin-manager workflow, or a shared template. Trigger here for "create a new plugin entry", "generated Claude and Codex manifests stay in sync", "move this maintainer workflow into plugin-manager", "marketplace source of truth", "plugin maintenance request", or "add a new skill folder to an existing plugin and update the plugin docs". Use plugin-health for audit-only conformance checks, plugin-release for final version bump / publish handoff, skill-improve for a bounded edit to one existing skill, and upstream-skill-harvest only for importing or refreshing a skill from an external upstream checkout.

Run plugin health, conformance, and packaging audits for one plugin or the whole marketplace. Use when checking plugin manifests, marketplace entries, skill frontmatter, required skill sections, routing fixtures, references, generated artifacts, or CI readiness. Inspired by GBrain skillpack-check and GStack health.

Release and upgrade workflow for marketplace plugins. Use when validating, versioning, documenting, publishing, creating a PR for, or preparing upgrade notes for plugin changes. Trigger here for "prepare a plugin release with version bump, render, validation, and upgrade notes", "ship this plugin change after confirming generated manifests are current", "bump the plugin version in marketplace.yaml and rerender both harness manifests", or "before we push and merge, verify release readiness and approval gates". Use manage-plugins for ordinary plugin maintenance or skill-folder creation before a release exists, and plugin-quality-gate only for second-opinion review, not release execution. Borrows GStack ship/document-release discipline and GBrain upgrade hygiene.

One-pass improvement workflow for an existing SKILL.md. Use when the user asks to improve, tighten, repair, simplify, deconflict, or upgrade a skill without running a full SkillOpt training session. Applies SkillOpt lessons such as evidence-first review, failure/success separation, bounded edits, validation, and rejected-change notes. Trigger here for "improve this SKILL.md in one bounded pass", "tighten the trigger text and output format for this skill", "repair this skill after a failed plugin-health check, but keep the edit small", or "make this skill better using SkillOpt lessons but do not create train selection or test splits". This does not create train/selection/test splits or run an offline optimization loop; use skillopt-training-run for repeated rollouts and held-out gates, and plugin-health for audit-only checks.

Capture scored rollout evidence for a SkillOpt-style skill optimization run. Use when building train, selection, or test trajectories for plugin skills, collecting task outputs, verifier results, tool-use summaries, failure modes, and before/after evidence for candidate skills.

Plugin-manager maintenance workflow for importing, adapting, diffing, and periodically resyncing skills from upstream GBrain or GStack checkouts into plugins in this marketplace. Use when harvesting a skill into plugins/<plugin>/skills, refreshing a plugin skill from upstream, auditing source drift, or recording source/adaptation notes. Trigger here for "harvest this GStack skill", "refresh a plugin skill from $GBRAIN_ROOT while preserving frontmatter and ledger notes", "diff our plugin copy against the upstream skill", or "import this external SKILL.md but scrub private paths and add routing fixtures". Use manage-plugins for newly authored local skills that do not come from upstream. This skill ships in the plugin-manager plugin because it manages other plugins rather than serving one plugin's end users.

Use this skill when the user has an existing PRD, product spec, one-pager, or PR/FAQ and wants critique, approval-readiness review, quality-bar scoring, required changes, or a red-team pass. Trigger on phrases like "review this PRD", "critique my product spec", "is this PRD ready", "PRD quality bar", "red-team this requirements doc", "find gaps in this PR/FAQ", or "what would block approval". Do NOT trigger for drafting a new PRD from an idea (use prd-writer), converting an approved PRD into an agentic implementation plan (use prd-to-plan), building a roadmap (use roadmap), or ranking multiple initiatives (use prioritization).

Use this skill whenever the user wants to write, draft, create, or scaffold a PRD (Product Requirements Document), spec, product brief, one-pager, or "working backwards" PR/FAQ from an idea, rough notes, discovery inputs, or an existing partial draft that needs authoring. Trigger on phrases like "write a PRD", "draft a spec", "product doc for X", "one-pager", "PR/FAQ", "feature brief", or when the user describes a feature/product idea and asks for it to be turned into a written requirements document. Do NOT trigger for critique or approval-readiness review of an existing PRD (use prd-review), engineering design docs, RFCs, ADRs, prioritization, roadmaps, or technical architecture documents. The output of this skill is always a markdown file (or section thereof) that a real product team could ship from.

Use this skill when the user wants to rank, score, triage, or compare product opportunities, features, PRD requirements, backlog items, roadmap candidates, or initiatives using RICE, impact-effort, confidence, reach, value, effort, risk, or tradeoff analysis. Trigger on phrases like "prioritize these features", "rank this backlog", "RICE score", "impact effort matrix", "what should we build first", "compare these initiatives", or "stack rank product bets". Do NOT trigger for a time-based roadmap (use roadmap), writing or reviewing one PRD (use prd-writer or prd-review), turning a PRD into an agentic plan (use prd-to-plan), or analyzing experiment results (use ab-testing or experimentation skills).

Use this skill when the user wants a product roadmap, initiative sequencing, Now/Next/Later plan, quarterly roadmap, release theme map, product strategy-to-roadmap translation, or executive roadmap narrative. Trigger on phrases like "build a roadmap", "roadmap these initiatives", "what should be now next later", "sequence this backlog", "quarterly product roadmap", "theme-based roadmap", or "plan the next two quarters". Do NOT trigger for drafting or reviewing a single PRD (use prd-writer or prd-review), converting a PRD into an agentic PLAN.md (use prd-to-plan), or scoring/ranking a set of options without a time horizon (use prioritization).

Use when the user mentions CLV, LTV, customer lifetime value, customer value prediction, lifetime revenue, CLV:CAC ratio, BG/NBD, Gamma-Gamma, RFM summary, purchase frequency prediction, monetary value prediction, churn probability, customer retention modeling, expected transactions, customer-level forecasting, or high-value customer identification. Also trigger on 'which customers are most valuable' or 'predict future customer revenue.' For grouping customers into segments or cohort retention curves use audience-segmentation; for lead propensity scoring use crm-lead-scoring; this skill predicts per-customer future value. If transaction data is not yet in the workspace, run data-extraction first.

Use when the user mentions social media analytics, social performance, Facebook insights, Instagram analytics, LinkedIn analytics, TikTok analytics, YouTube analytics, X analytics, Twitter analytics, social engagement, social reach, share of voice, social sentiment, brand mentions, social content performance, viral content, social ROI, social listening, or social benchmarking. Also trigger on 'how are we doing on social' or 'what's performing on LinkedIn.' For a full cross-channel competitor benchmark (keywords, ads, pricing, traffic) use competitive-intel; for paid-ad platform metrics like CPA/ROAS use paid-media; this skill owns owned/earned social performance and sentiment. If social platform data is not yet in the workspace, run data-extraction first.

Developer-experience review for plugin onboarding. Use when evaluating a plugin's fresh-clone setup, README flow, local testing instructions, prerequisites, bundled CLIs, confusing paths, or contributor experience. Inspired by GStack devex-review and QA workflows.

Use when the user wants to analyze A/B test or experiment results and needs significance, confidence intervals, effect sizes, lifts, or a readout report. Triggers on phrases like "analyze this test", "run the readout", "is this significant", "what's the lift", "did B win", or any time the user shares an experiment dataset, summary table, or readout request. Accepts data as pasted CSV/TSV, a summary table (variant / visitors / conversions), a file path, or raw numbers in prose. Produces a reproducible analysis directory, a structured numerical summary, and a plain-English interpretation with diagnostic flags.

Use when the user is starting a new A/B test, formalizing a half-formed test idea, writing a test plan or PRD, or pre-registering an experiment. Triggers on phrases like "design a test", "set up an experiment", "write the test plan", "pre-register this", "what should we measure", "what's a good hypothesis", or any time the user describes a product/marketing change they want to validate experimentally. Produces a versioned, pre-registered experiment design document with hypothesis, OEC, guardrails, counter-metrics, randomization design, power/MDE estimate, pre-specified analysis plan, and operational decision criteria — written to a reproducible artifact directory the auditor and readout tooling can later reference.

Use when the user wants a stakeholder-ready experiment readout, report, or summary written from analyzed results. Triggers on phrases like "write the readout", "generate the experiment report", "summarize this test for leadership", "draft the post-test writeup", "exec summary of the test", "share this with the team". Produces a structured Markdown report tuned to a named audience (executive / technical / mixed / regulatory), grounded in the pre-registered design and the analysis artifact, with a single explicit recommendation. Does not invent numbers — it composes from the design document, the analysis directory, and the user's stated recommendation context.

Use when the user wants a code review of A/B test implementation — assignment logic, bucketing, feature flags, exposure event logging, variant propagation, and metric instrumentation. Triggers on phrases like "review this experiment code", "is my bucketing correct", "check the implementation before launch", "review the flag logic", "audit the assignment code", "why might this SRM", or any time the user shares a PR, file, or directory containing experiment plumbing. Produces a severity-tagged findings report with file:line citations and concrete fix suggestions, written to a reproducible artifact directory. Complements (not replaces) the `experiment-auditor` sub-agent: this skill reviews the *code*, the auditor reviews the *design and the readout*. Most production SRM is caused by bugs this skill is built to find.

Use when the user needs to calculate the required sample size, duration, or minimum detectable effect (MDE) for an A/B test. Triggers on phrases like "how many users do I need", "what sample size", "calculate MDE", "how long should this test run", "do we have enough traffic", "is this test feasible", "power calculation", or any planning question that involves trading off effect size, sample, and duration. Produces a reproducible sample-size artifact with the headline n-per-arm + duration, a two-direction sensitivity table (n at multiple MDEs, MDE at multiple durations), adjustments for CUPED / cluster designs / multiple comparisons / sequential testing, and the MDE translated into business units. Delegates to the `experiment-statistician` sub-agent for ratio metrics, heavily skewed continuous metrics, and cluster-randomized designs that require simulation.

Analyze complex experiments whose design breaks ordinary two-proportion or t-test analysis: sequential testing, Bayesian decision rules, CUPED/CUPAC, ratio metrics, correlated or clustered observations, CATE, uplift modeling, and bandit-versus-controlled tradeoffs. Trigger when the analysis method itself is the hard part, including fitting an uplift model or estimating CATE. For a standard significance or lift readout, use ab-testing analyze-results; for whether subgroup evidence is safe to deploy, use personalization-governance.

Review experiments for financial-services compliance, conduct risk, model risk, fairness, disclosures, advice-versus-marketing boundaries, dark patterns, trust erosion, auditability, and consumer perception. Trigger before customer-facing tests in regulated or high-trust contexts. For regulatory screening of published advertising or marketing materials rather than experiments, use marketing-analytics compliance-review instead.

Monitor early experiment health and temporal signal maturity: first-48-hours reads, sample-ratio- mismatch (SRM), exposure validation, operational guardrails, novelty and primacy effects, proxy signals, and early-versus-late evidence. Trigger when interpreting opening-day results. For the assignment or logging bug behind an SRM, use ab-testing review-experiment; for the final ship call, use experiment-decision-review.

Design or evaluate email experiments for true incremental lift: post-Apple-MPP measurement, holdout and control-group design, send-frequency and fatigue effects, proxy metrics, delayed outcomes, and financial-services email constraints. Trigger when the question is whether email caused the outcome — incrementality, holdout, or causal lift. For day-to-day email campaign performance rather than causal lift, use marketing-analytics email-analytics instead.

Convert experiment evidence into a calibrated, board-ready executive decision memo — uncertainty, assumptions, and limitations intact — when an experiment result will justify launch, budget, compliance, or leadership action. Trigger for board-ready decision memos and leadership readouts built from experiment evidence. For a general stakeholder experiment report from an analysis directory, use ab-testing experiment-report.

Review existing experiment evidence and recommend ship, kill, iterate, extend, retest, or escalate using decision thresholds beyond statistical significance — business materiality, risk, guardrails, and temporal validity. Trigger when a result exists and the call is contested: significant but small, flat but useful, early but tempting, or positive with a guardrail regression. For computing significance or lift, use ab-testing analyze-results; to package the decision for leadership, use executive-evidence-brief.

Design the standing experimentation operating model — center of excellence, review board, decision rights, maturity curve, first-year roadmap, governance process, and earned-autonomy model. Trigger when the problem is repeatable organizational experimentation capability, not one test. For designing a single experiment, use safe-experiment-design or ab-testing design-experiment instead.

Reconcile experiment evidence with other measurement systems — MMM, attribution, geo-lift, global holdouts, and proxy calibration — when systems disagree and a budget or spend decision depends on the resolution. Trigger when experiment and non-experiment measurement conflict and need calibrating. For building or tuning that measurement model itself, use marketing-analytics attribution-analysis.

Capture null, flat, negative, and inconclusive experiment results into a durable experiment- learnings registry so failed and ambiguous tests compound into institutional memory instead of being silently rerun — repository schemas, tagging, learning summaries, and meta-analysis. Trigger after a flat, invalid, negative, or ambiguous experiment when the learning must be preserved for reuse. This is an experiment-results archive, not a document or note store.

Decide whether personalization, segmentation, CATE, or uplift evidence is strong and safe enough to deploy in high-trust or regulated environments — the governance and deployment call, not the modeling. Trigger when a team wants to target, suppress, personalize, or automate based on subgroup experiment results. For the underlying subgroup modeling itself, use advanced-experiment-analysis.

Plan experiment feasibility and program-level duration strategy in regulated or low-velocity settings: is a test feasible given scarce traffic, what MDE is realistic, is a default 30-day habit justified, and what low-velocity alternatives (validated proxies, sequential monitoring, non- experimental paths) apply when outcomes are delayed or stakeholders want to stop early. Plan a defensible duration as a decision-latency and risk tradeoff. For the hands-on sample-size math for one A/B test, use ab-testing sample-size instead.

Design safe, governed experiments for regulated or high-trust settings: hypotheses, success metrics, guardrails, pre-registration, compliance-constrained and pre-registered plans, risk tiers, kill switches, and audit trails for a regulated journey. Trigger when a test touches investor-facing content, financial-services journeys, personalization, or any irreversible customer exposure. For an ordinary product test without regulatory stakes, use ab-testing design-experiment; for the standing program, use experiment-operating-model.

Orchestrate a long KB operation — migration, archive scan, enrichment batch, research backfill — as batched, checkpointed, resumable work with sample-before-bulk gates, and save/restore resumable context so a job survives across sessions. Trigger on 'run this KB job in batches', 'long/background KB job', 'save/resume KB progress', 'checkpoint this job'. Absorbs context-checkpoint. Recurring-schedule design lives in references/automation.md; the actual ingestion is done by ingest or media-ingest.

Assemble a proactive briefing from KB context — a daily digest, meeting prep, or a project/entity brief — surfacing risks, decisions needed, open loops, changed state, and source links. Trigger on 'brief me on X', 'daily briefing', 'meeting prep', 'prep me for tomorrow's meeting', 'what should I know before'. This pushes a prepared briefing document; to pull the answer to one specific question use query; for a saved, timestamped analytical report with a source manifest use reports.