Path kinds

guided

A structured sequence with a fixed order and gated steps.

self-paced

The same content without gating; the learner picks the order.

interactive

Quizzes and simulations against sample workspace data.

hands-on

Every step is a real task in the learner's own workspace.

micro-learning

Bite-sized units surfaced inline next to a finding.

case-study

A completed analysis walked through end to end.

playbook

A step-by-step operational guide promoted from knowledge.

• — A path declares the skills it moves; an unclaimed path cannot be published.
• — Steps unlock in order for guided paths and never for self-paced ones.
• — A path whose lesson was retired is superseded, not silently shortened.

Lesson blocks

Lesson states

• — A lesson without at least one objective and one skill is invalid.
• — Theory is optional; practice is not — reading alone never completes a lesson.
• — A lesson renders through the content-engine renderer, never a bespoke viewer.
• — Lessons load lazily and their progress is cached locally for offline reading.

Mastery levels

Categories

Rules

• — A skill level only rises on validated evidence: practice, challenge or a real workspace task.
• — Evidence is append-only; a level can decay over time but its history is never rewritten.
• — A skill with unmet dependencies cannot exceed intermediate.
• — Extensions contribute skills through the SDK and own them in the registry.

Practice domains

support

Fix a real finding: metadata, broken links, accessibility violations.

analytics

Interpret connected GA4 and Search Console data and draw a conclusion.

extensions

Install, configure and run an extension against the workspace UWO.

marketplace

Discover, install or publish an asset and verify its entitlement.

content

Draft, version and publish a knowledge object through the content engine.

connectors

Connect a provider, run a sync and read the normalized envelope.

• — Practice runs against the learner's own workspace data, never a mock fixture, unless the workspace is empty.
• — A practice run is a validated command: it obeys permissions, policy, credits and approval.
• — Validation is objective and re-runnable; a self-declared completion is never accepted.
• — A destructive exercise requires explicit confirmation and offers a rollback.

Challenge scopes

daily

Review one open finding and record the decision.

weekly

Optimize five pages and re-run the snapshot to confirm the delta.

mission

Advance one mission objective by a measurable step.

extension

Build and install a custom connector or extension.

community

Peer review a published asset — planned.

Streaks

• — A streak counts days with at least one validated learning or practice event.
• — A missed day breaks the streak; the longest streak is kept as a record.
• — Rewards are credited through the monetization wallet, never minted locally.

Capability registry

learning.lessonlow · 0 cr

Define, version and serve a lesson.

learning.pathlow · 0 cr

Compose lessons into a role- or mission-aligned sequence.

learning.practicemedium · 2 cr

Run a hands-on exercise against workspace data.

learning.challengelow · 0 cr

Offer, accept and settle a challenge.

learning.quizlow · 0 cr

Score an assessment and record the result.

learning.skilllow · 0 cr

Read and update skill mastery from validated evidence.

learning.certificatehigh · 1 cr

Issue or revoke a credential with its evidence hash.

Permissions

learning.read

Read learning content, own progress and the skill graph.

learning.participate

Take lessons, quizzes and challenges.

learning.practice

Run practice exercises that write to the workspace.

learning.author

Create and version lessons, paths and skills.

learning.certify

Issue and revoke credentials.

learning.admin

Configure missions, policies and workspace-wide learning settings.

Rules

• — Learning capabilities register in the same kernel registry as every other capability.
• — The AI planner discovers them dynamically; there is no learning-specific intent table.
• — A capability the workspace cannot run is filtered out before planning.
• — Everything registered is owner-attributed and versioned.
• — Discovery is a filter over the registry; there is no hardcoded catalogue.
• — A retired entry stays resolvable so issued credentials keep their evidence.

Widgets

Continue learninglearning.read

The next lesson or challenge from the coach, with its estimated time.

Mission progresslearning.read

Objectives, current metric and the paths attached to each objective.

Skill growthlearning.read

Recent level changes and the weak areas the workspace actually needs.

Recent activitylearning.read

Completed lessons, solved challenges and issued credentials.

Streaklearning.read

Consistency tracking with the current and longest streak.

Badges and certificateslearning.read

Issued credentials with their evidence links.

Context dock

current-lesson

The lesson attached to the surface the user is on.

required-skills

Skills the current task needs and the learner's level in each.

related-knowledge

Knowledge objects that explain the active finding.

time-to-complete

Estimated minutes for the recommended next step.

Rules

• — The learning dashboard is a widget surface resolved from the same dashboard registry as every other surface.
• — Widgets are permission-filtered before render; nothing is shown disabled.
• — The dock is read-only and reflects live progress, never a cached estimate.
• — Progress syncs incrementally and can always be force-refreshed by the user.

Consumed events

report.generated

Attach a lesson and knowledge object to every open finding.

finding.resolved

Validate the matching practice and raise the related skill.

snapshot.completed

Recompute mission objective metrics and path relevance.

connector.sync.completed

Unlock analytics practice that needs live provider data.

asset.installed

Register learning content contributed by the installed asset.

ai.plan.created

Surface required skills for the plan in the context dock.

Emitted events

Report integration

• — Every finding links to a lesson or guide, a knowledge object and an AI explanation.
• — The link is resolved from the finding type, never hardcoded per report.
• — Community discussion on a finding is planned, not shipped.

Marketplace contributions

learning-pack

Courses and paths published as an installable asset.

challenge-pack

A set of challenges scoped to a role or mission.

certification-pack

Credentials with their criteria and evidence rules.

knowledge-pack

Guides, playbooks and templates lessons reference.

Learning SDK

defineLesson(lesson: Lesson) => LessonBinding

Register a versioned lesson with objectives, skills and practice.

defineSkill(skill: Skill) => SkillBinding

Declare a skill, its category and its dependencies.

definePath(path: LearningPath) => PathBinding

Compose lessons into a role- or mission-aligned sequence.

definePractice(exercise: PracticeExercise) => PracticeBinding

Bind a hands-on exercise to a kernel capability and validation.

defineChallenge(challenge: Challenge) => ChallengeBinding

Publish a scoped challenge with a measurable target.

defineCredential(credential: Credential) => CredentialBinding

Declare a badge or certificate and its criteria.

defineMissionTemplate(mission: Mission) => MissionBinding

Ship a reusable mission with objectives and attached paths.

defineCoachSource(source: RecommendationSource) => CoachBinding

Contribute recommendations to the AI learning coach.

defineLesson({
  id: "seo.metadata.essentials",
  title: "Metadata Essentials",
  objectives: ["Fix missing metadata"],
  skillIds: ["seo.content"],
  practiceId: "practice.metadata.fix",
});

defineSkill({
  id: "seo.metadata",
  category: "seo.content",
  masteryLevels: ["beginner", "intermediate", "expert"],
});

• — The SDK is the only way an extension contributes learning content.
• — Every definition is versioned and owner-attributed in the registry.
• — No SDK function can issue a credential without declared criteria or skip validation.
• — SDK versions follow the platform version and stay backwards compatible.

Performance

• — Lessons and challenges load on demand; the catalogue is never fetched whole.
• — Completed items are cached locally so progress reads work offline.
• — Progress updates sync incrementally in the background.
• — A manual refresh always forces a full re-sync.

Security and compliance

• — Learning progress and analytics are user-owned; AI recommendations require explicit consent.
• — Extension-contributed learning runs sandboxed under declared permission scopes.
• — Marketplace learning content passes automated checks plus human review.
• — Credential records are immutable and audit-trailed for GDPR and SOC 2 evidence.