Skills
Audit a non-signup form with ak:form-cro
Review form friction, data necessity, accessibility, measurement, and test hypotheses before approving implementation.
Use ak:form-cro to audit a lead-capture, contact, demo-request,
application, survey, checkout, or quote-request form. It turns the current flow
and evidence into prioritized findings, a proposed field design, copy and error
guidance, and test hypotheses. Recommendations are not live form changes.
Choose ak:form-cro for a non-signup form
Use ak:form-cro when
- You need to understand form starts, completion, field errors, abandonment, or mobile differences.
- You want to justify every required and optional field against the follow-up process and compliance needs.
- You need recommendations for layout, labels, validation, multi-step flow, submit states, or post-submit expectations.
Choose another workflow when
- The form creates an account. Use the signup-flow CRO workflow.
- The form sits inside a popup and the container behavior is the main issue. Use the popup CRO workflow.
- The surrounding page, offer, or traffic quality is the primary problem. Audit that broader surface before attributing the result to the form.
- You already selected a testable change. Use
ak:ab-test-setupto design the experiment.
Prepare the form evidence
Provide the form URL, screenshots, or field list; the form type and audience; what happens after submission; current starts and completions; field-level errors and drop-off; device split; and the fields actually used downstream. Also provide accessibility, consent, privacy, retention, security, and legal requirements. Remove secrets and unnecessary personal records.
Complete Onboarding, and confirm Marketing Kit is installed for the runtime and scope you are using.
| Runtime | Invocation | Availability boundary |
|---|---|---|
| Claude Code | /ak:form-cro ... | Native delivery is the default; explicit plugin delivery is also supported. |
| Cursor | /ak:form-cro ... | Slash invocation is user-verified; broader runtime parity is not implied. |
| Codex | $ak:form-cro ... | The Skill uses native Codex discovery; Hook projection is partial. |
See Runtime adapters for component-level differences.
Run the Skill
/ak:form-cro "Audit this demo-request form using the supplied field analytics. Minimize personal data, preserve required consent and accessibility, propose tests, and do not edit or publish the form"/ak:form-cro "Audit this demo-request form using the supplied field analytics. Minimize personal data, preserve required consent and accessibility, propose tests, and do not edit or publish the form"$ak:form-cro "Audit this demo-request form using the supplied field analytics. Minimize personal data, preserve required consent and accessibility, propose tests, and do not edit or publish the form"Understand the review stages
- Classify the form. The Skill identifies the form's purpose, audience, completion outcome, follow-up process, and business constraints.
- Establish evidence. It reviews starts, submissions, field drop-off, validation errors, completion time, and device differences without inventing missing baselines.
- Audit fields and flow. It challenges field necessity, required status, order, grouping, conditional logic, multi-step behavior, and progress saving.
- Review interaction quality. It covers persistent labels, useful help, keyboard and autofill behavior, accessible errors, preserved input, loading, success, and failure states.
- Prioritize recommendations. Each finding receives a rationale and priority; any expected conversion effect remains a hypothesis, not a promise.
- Define measurement. The Skill proposes view, start, field, error, submit, and successful-submission events plus bounded experiments.
Keep consent and implementation separate
Less friction must not mean less informed choice
Do not remove required disclosures, preselect consent, hide optional status, misrepresent urgency, or make refusal harder than submission. Conversion does not override privacy, accessibility, security, or legal obligations.
- Ask only for data that has a documented purpose, lawful basis where required, retention rule, and owner.
- Do not infer or enrich sensitive attributes without approval and a valid data policy. A provider name in a recommendation does not authorize network access or data transfer.
- Keep labels visible, errors specific, input preserved, and keyboard-only and assistive-technology paths usable.
- Treat phone numbers, free-text messages, payment details, and application data according to their sensitivity, not as ordinary conversion fields.
- Approve code edits, analytics changes, vendor configuration, publishing, and production rollout separately from the audit.
Review the outputs and evidence
A complete result should include:
- A form audit with finding, evidence, impact rationale, recommendation, and priority for each issue.
- A justified required/optional field list, order, conditional behavior, labels, help text, error copy, submit copy, and post-submit states.
- Mobile and accessibility considerations.
- An event plan that avoids capturing field values unless explicitly required.
- Test hypotheses with primary metrics, quality guardrails, and assumptions.
- A clear boundary between recommendations and any separately approved implementation work.
Troubleshoot the workflow
| Symptom | Safe next step |
|---|---|
| No field-level evidence exists | Add privacy-reviewed start, error, submit-attempt, and success events before ranking fields by impact. |
| More leads arrive but quality falls | Use downstream qualification as a guardrail and review whether required context was removed. |
| A field is legally required | Keep it, explain why it is needed, and test clarity or timing rather than silently deleting it. |
| Mobile abandonment is high | Verify keyboard type, autofill, tap targets, viewport, validation, and data preservation on real devices. |
| The runtime cannot find the Skill | Confirm target and scope, restart the runtime, then follow Runtime cannot find a Skill or Agent. |
Know the limits
- Field-count and layout rules are hypotheses, not universal effect sizes.
- Completion rate alone does not measure lead quality, user understanding, successful payment, or downstream value.
- The Skill cannot verify live accessibility, analytics, compliance, or vendor behavior without the corresponding access and evidence.
- The documented package releases contain identical
ak:form-crocontent and invocations.
Continue with ak:ab-test-setup after choosing a recommendation, or review the
Marketing Kit overview.
Design, analyze, or optimize a funnel with ak:funnel
Map funnel stages and metrics, audit observed drop-off, or produce prioritized optimization and test recommendations.
Improve post-signup activation with ak:onboarding-cro
Audit first-run friction and design an evidence-led onboarding flow, copy, measurement plan, and bounded experiments.