| name | discovery-onboarding |
| description | collect and persist reusable lightspeed website discovery defaults in memory before discovery workflows. use when a lightspeed teammate starts a client website discovery task, asks to create a discovery pack or summary, or a related discovery skill is blocked by missing client, website, goal, output, approval, or standing project defaults. route away from this skill for run-specific source gathering, draft review, specialist audits, launch planning, or one-off rewrites when enough context already exists. |
Discovery Onboarding
Purpose
Use this skill to collect only the durable defaults that help LightSpeed teammates run future website discovery work without repeating the same setup questions.
This skill is an onboarding gate, not a discovery pack generator. Its job is to decide whether reusable defaults are missing, save only those defaults, and then resume or route the original request.
Supporting files
references/onboarding-contract.yaml — source of truth for required fields, optional fields, trigger rules, skip rules, and memory keys.
references/related-skill-routing.md — routing map for the wider LightSpeed discovery skill set.
Always read references/onboarding-contract.yaml before asking onboarding questions. Read references/related-skill-routing.md when the request overlaps with another discovery skill, asks which workflow to use, or appears too specialist for onboarding.
Memory state
Use {{label:Memory,id:file_persistence,type:file_persistence}} as the backing store for the onboarding keys in references/onboarding-contract.yaml.
Persist defaults for the current runtime end user in discovery-user-defaults.yaml, and read that file before asking the user to restate defaults.
Store only durable defaults defined in the onboarding contract. Do not store one-off project notes, meeting detail, stakeholder quotes, support findings, audit observations, draft conclusions, or temporary source evidence.
Trigger decision
Run this skill only when all three conditions are true:
- The request is a LightSpeed website discovery, website audit, discovery pack, or discovery-readiness task.
- The current request or downstream discovery workflow needs reusable defaults such as
client_name, website_url, primary_goal, default_output_mode, preferred_template, approval context, or standing project preferences.
- Those reusable defaults are missing from both the current request and
discovery-user-defaults.yaml.
Skip this skill when:
- the user provided enough context to complete the task now
- the request is only a rewrite, cleanup, summary, or formatting pass
- the request is about run-specific evidence intake from notes, links, transcripts, files, emails, issues, or feedback
- the request is reviewing an existing discovery output
- the request needs a specialist discovery assessment rather than reusable defaults
- asking onboarding questions would block useful partial progress unnecessarily
Routing rules
Treat onboarding as a lightweight prerequisite. Do not let it capture work that belongs to a specialist skill.
Route according to references/related-skill-routing.md:
- Use this skill first only when missing reusable defaults would make the specialist workflow unsafe or repetitive.
- Use
discovery-source-intake when the task needs current-project evidence gathered or normalised.
- Use
discovery-pack-review when a draft discovery pack, section, or client summary already exists and needs QA.
- Use the specialist reviewer skills for hosting, performance, content, email lists, and security when the user asks for those assessments.
- Use broader LightSpeed project intake, PRD, launch, QA, or implementation skills when the request has moved beyond discovery.
When both onboarding and another discovery skill apply, do this sequence:
- Extract and save any reusable defaults already present in the request.
- Ask only for the first missing required default if the downstream task cannot proceed safely without it.
- Continue into the specialist skill as soon as the minimum required defaults are present.
Preflight workflow
- Read
references/onboarding-contract.yaml.
- Read
discovery-user-defaults.yaml from Memory if it exists.
- Check the current request against the contract trigger, skip, and route rules.
- Identify any contract-defined defaults already supplied by the user.
- Save supplied durable defaults under the exact
store_as keys from the contract.
- Re-check required state after saving request-supplied values.
- If required state is present, continue the original request or route to the appropriate related skill.
- If required state is missing, ask only the smallest useful onboarding batch.
Input handling rules
Partial inputs
If the user gives partial project context, save what is durable and ask only for the missing required field that blocks the current task.
Example:
- User provides client name and goal but no URL.
- Save
client_name and primary_goal.
- Ask only for
website_url, unless the task can proceed without it.
Multiple clients or websites
If the user mentions multiple clients, sites, domains, brands, or publications:
- do not merge them into one memory record
- ask which client/site should be treated as the active discovery default
- save shared network-level defaults only when the user clearly presents them as reusable across the whole programme
Ambiguous source material
If the user supplies notes, links, transcripts, attachments, emails, Asana items, GitHub issues, or feedback:
- treat those as run-specific evidence
- route to
discovery-source-intake after saving any clearly durable defaults
- do not store the evidence itself in Memory
Existing drafts
If the user provides or references an already assembled discovery pack, section, client summary, or follow-up document:
- route to
discovery-pack-review
- use onboarding only if reusable project defaults are missing and needed for the review
Specialist discovery areas
If the request is primarily about hosting, performance, content, email list quality, security, technical SEO, analytics, redirects, claims, policy/governance, or launch readiness:
- use the relevant specialist skill directly
- run onboarding only if the specialist workflow needs missing reusable project defaults
Output preference
If the user does not specify output mode, default to LightSpeed internal working output.
Use client-facing wording only when the user explicitly asks for a client-ready, client-facing, shareable, proposal-ready, or stakeholder-facing document.
Onboarding batches
Ask in small batches, not as a long intake form. Treat each batch as one concise question that covers closely related fields.
Batch 1: project basics
Ask this first only when required state is missing.
Collect only:
client_name
project_name when the user distinguishes it from the client name
website_url
primary_goal
Batch 2: business and scope context
Ask only when the current task would materially benefit from standing setup details.
Collect when useful:
business_summary
target_audience
key_pages
required_features
content_needs
integrations
Do not block the current task on this batch unless the missing detail prevents a correct output.
Batch 3: delivery and approval context
Ask only when standing delivery assumptions affect the requested output.
Collect when useful:
timeline
budget_range
stakeholders
decision_process
discovery_lead
Do not block the current task on this batch unless approval or delivery context is essential.
Batch 4: output and template preferences
Ask only when output preferences are not obvious and materially affect the current deliverable.
Collect when useful:
default_output_mode
preferred_template
pack_preferences
Use safe defaults when the user does not provide these preferences.
Completion rules
- Completion requires the required state from the contract, not every optional field.
- Once the required state is present, stop onboarding and resume or route the original request.
- Confirm only the defaults that matter for the current task.
- Do not summarise onboarding for its own sake.
- Do not ask the user to restate information that is already in the request, memory, or available source evidence.
Output rules
When onboarding runs, respond with one of these patterns:
- Small missing-field question — ask the minimum needed to proceed safely.
- Saved-default confirmation plus continuation — briefly confirm saved defaults and continue the requested task.
- Routing continuation — briefly state the chosen next workflow only when helpful, then continue.
When onboarding does not run, do not mention it. Continue the user's original request or route to the correct related skill.
LightSpeed team usability rules
- Prefer concise, practical prompts that a teammate can answer quickly.
- Keep internal and client-facing outputs separate.
- Preserve approval and evidence boundaries: confirmed, assumed, inferred, unresolved, internal note.
- Avoid generic agency discovery language when a LightSpeed-specific project, WordPress build, block theme, WooCommerce site, publishing site, tourism site, or AI-readiness workflow is implied.
- Ask no more than one focused question before making useful progress, unless the user explicitly asks for a full intake form.
Maintained by the 🤖 LightSpeedWP Automation Team
📋 AI Governance · 🧠 Agents · 📞 Contact