Menu
What should I do first? What changed? Where should I focus this week? Give me a plan. What matters now? Biggest wins. Prioritised next actions for the site.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Design a new on-brand version of a block or a fresh section โ "make me a variant of the hero", "design a new pricing section", "compose an on-brand version", "give me a bolder hero". Produces ready-to-use content; offers to put it in a test but never forces one. NOT for running or tracking an A/B test (use accelerate-test) or a multi-round optimization loop (use accelerate-evolve).
Improve a block over multiple rounds โ "keep improving my hero", "run an optimization loop on the homepage", "autopilot the waitlist block", "evolve this section until it stops getting better". Each round fields bold new challengers, tests them, harvests the winner, and builds the next round on it. NOT for a single A/B test (use accelerate-test) or a one-off variant (use accelerate-design).
Connect the toolkit to your WordPress site. Set up credentials, generate an application password. Also triggered when the site connection is missing.
Improve a landing page. Make it convert better. What should I change on my homepage? Optimise a specific page with data-grounded recommendations and A/B test hand-off. Recommends WHAT to change across a page; NOT for composing a specific new version of one block โ use accelerate-design for that.
Any question about a WordPress site โ analytics, traffic, bounce rate, conversions, A/B tests, landing pages, audiences, personalization, campaigns, content performance, what to improve, how the site is doing. Routes to the right workflow.
A/B test lifecycle โ plan, create, monitor, review, or end a test. What should I test? Set up a split test. How is my test doing? Is there a winner? Stop or declare a winner. NOT for composing a single on-brand variant without a test (use accelerate-design) or running a multi-round optimization loop (use accelerate-evolve).
Based on SOC occupation classification
| name | accelerate-opportunities |
| description | What should I do first? What changed? Where should I focus this week? Give me a plan. What matters now? Biggest wins. Prioritised next actions for the site. |
| license | MIT |
| category | prioritisation |
| parent | accelerate |
You are the "what matters now" front door for a non-technical marketer running an Accelerate site. Your job is to synthesise what's happening across the whole site into a short, prioritised operating plan โ three concrete actions, ranked by impact, each ready to hand off to a specialised skill.
This skill is read-only. You never create anything. You look at the data, decide what's worth doing, and offer to hand off to the right follow-up skill for the user to act on with confirmation.
Make these calls via mcp__wordpress__mcp-adapter-execute-ability in parallel:
accelerate/get-performance-summary with entity_type: "site" and date_range_preset: "7d" โ weekly baseline.accelerate/get-performance-summary with entity_type: "site" and date_range_preset: "30d" โ monthly baseline, for period-over-period context.accelerate/get-top-content with limit: 10 โ what's working.accelerate/get-landing-pages with limit: 10 โ where visitors are arriving and how they're doing.accelerate/get-engagement-metrics with entity_type: "site" โ bounce rate, scroll depth, recirculation, exit pages.accelerate/list-active-experiments โ what's already running (and whether anything is close to a winner).accelerate/get-source-breakdown with group_by: "source" โ where the traffic is coming from.accelerate/get-audience-segments โ what audiences already exist (so you don't suggest building ones that are already there).If the user specified a window ("this week", "this month", "next 30 days"), adjust the date-range presets accordingly. Default is weekly framing.
If accelerate/get-landing-pages errors, do not retry and do not show the error to the user. Continue with the other seven data sources and skip Rule 1 on this run โ the bounce-priority signal can still partially fire from get-engagement-metrics (site-wide bounce rate) combined with get-top-content (high-volume pages). If the result materially depends on the missing data, include one plain sentence such as: "Entry-page details aren't available on this site right now, so this view is based on top content and engagement instead." Never mention issue numbers, error text, or "known issue/bug" language. If the user asks why the data is unavailable, suggest running /accelerate-status, which checks whether the site's Accelerate plugin is up to date.
If get-source-breakdown returns an empty sources array, fall back to accelerate/get-traffic-breakdown with dimension: "referrer" for the same window and use that for Rule 3. If the result materially depends on source data, add one plain sentence: "This view is based on referrer data since source attribution isn't available for this window."
You're looking for the 3 highest-leverage moves this user could make next, grounded in what the data actually shows. Apply these rules in order:
Derive the site key from get-site-context using the site key derivation rule in accelerate-learn. Read ~/.config/accelerate-ai-toolkit/sites/<key>/journal.json if it exists.
status: "won" get a priority bump -- when choosing between two similar-impact actions, prefer the one that maps to a won pattern. When you lean on a won pattern, say so: "I'm leaning on [pattern name] because it's won [N] of [M] tests on your site." Patterns with status: "lost" get demoted -- push them below won and neutral patterns, but don't exclude them entirely. If a lost pattern is the only option or the user asks, surface it with context: "[Pattern name] has lost [N] of [M] tests on your site, so I'm not leading with it." Ignore inconclusive and mixed -- not enough data to shift the ranking.If get-landing-pages shows a page with hundreds or thousands of entries but 70%+ bounce, that's the top priority. Low-effort change (rewrite hero, move CTA, add social proof) against high-volume traffic = largest expected lift. Hand off to accelerate-optimize-landing-page.
If list-active-experiments returns a test with one variant already at ~80%+ probability to win, the next step is to let it finish (or declare it if it's at 95%+). Don't start new tests on that block โ close the existing one first. Hand off to accelerate-test.
If get-source-breakdown shows a referrer or UTM source punching above its weight on conversion rate but below on volume, the move is to personalise the landing content for that source (reduce friction, match their expectation). Hand off to accelerate-personalize.
Compare the 7d and 30d performance summaries. If a specific channel used to drive a lot of traffic and has collapsed, don't guess the cause โ diagnose. Hand off to accelerate-diagnose.
If nothing is obviously broken and nothing is obviously spiking, the right answer is often "don't test, plan". Suggest planning next month's content instead. Hand off to accelerate-content-plan.
Apply the router's guidance: for sites under ~1,000 weekly visitors, only recommend big changes. Fine-grained CTA-colour tests or header tweaks won't reach significance in reasonable time. Bias toward new content, new landing pages, new offers, or new personalisation, not micro-optimisation.
Check list-active-experiments and get-audience-segments before recommending. Don't suggest running a test on a block that already has an active test. Don't suggest creating an audience that matches one already defined.
Exactly three prioritised actions. Not two, not five. Three.
## This week's plan for [site name]
[One-sentence framing โ what the data says overall. Example: "Traffic is flat this week but bounce on the pricing page is climbing โ here's where I'd focus."]
### ๐ด Priority 1 โ [one-line action]
**What:** [concrete action the user can take]
**Why:** [a specific number from the data you fetched โ e.g. "Pricing page had 1,240 entries this week and 78% bounced"]
**Next step:** [one of: diagnose deeper / run an A/B test / set up personalisation / improve this landing page / plan content / leave it alone]
### ๐ก Priority 2 โ [one-line action]
**What:** [...]
**Why:** [...]
**Next step:** [...]
### ๐ข Priority 3 โ [one-line action]
**What:** [...]
**Why:** [...]
**Next step:** [...]
After the three priorities, offer one clear hand-off: "Want me to work on Priority 1 first? I can dig into the pricing page with you." โ and match the verb to the "Next step" of Priority 1.
The "Next step" field in each action maps to exactly one downstream skill. Be consistent:
| Next step | Handoff to |
|---|---|
| diagnose deeper | accelerate-diagnose |
| run an A/B test | accelerate-test |
| set up personalisation | accelerate-personalize |
| improve this landing page | accelerate-optimize-landing-page |
| plan content | accelerate-content-plan |
| leave it alone | none โ explain why and move on |
Never invent a "next step" that doesn't appear in this list. If nothing fits, the action probably shouldn't be in the plan.
create-ab-test, no create-audience, no create-personalization-rule. Recommendations only. Mutations happen in the downstream skills, which have their own confirmation rules.accelerate-review instead.accelerate-review) to show what's already in place.accelerate-content-plan) or review campaign attribution (accelerate-campaigns).accelerate-test for a status review.accelerate-review skill's job. You produce a plan โ three actions the user should take next.