Progression Frontier, Prerequisite Evidence Gap and Exploration Projection v1¶

Status: PR8 implementation contract

Outcome¶

PR8 introduces Capability Lab's first deterministic person-scoped advisory progression projection. It composes explicitly selected governed capability-state dimensions, direct shared capability relations, explicit request-local focus, and explicit preserved exploration into inspectable frontier candidates, prerequisite evidence gaps, and exploration opportunities.

It does not infer goals, whole-person capability, readiness, priority, difficulty, optimal paths, safety, permission, professional authority, or human value.

PROGRESSION FRONTIER != RECOMMENDATION
FRONTIER CANDIDATE != NEXT REQUIRED STEP
FRONTIER CANDIDATE != READINESS
FRONTIER CANDIDATE != PERMISSION

Layer boundary¶

CapabilityCatalog
      +
explicitly selected PersonalCapabilityState dimensions
      +
explicit ProgressionFocus
      +
explicit ExplorationInput
      |
      v
core:deterministic_progression_frontier@1
      |
      +--> FrontierCandidate
      +--> PrerequisiteEvidenceGap
      +--> ExplorationOpportunity
      |
      X
no claim/state/history/permission mutation

PR8 reads already-governed state representation. It may validate the selected state's exact PR1/PR2/PR3 basis, but it does not reinterpret raw evidence or create new personal capability state.

PR8 READS GOVERNED STATE
PR8 DOES NOT RE-EVALUATE RAW EVIDENCE
PR8 DOES NOT DERIVE PERSONAL CAPABILITY STATE

Explicit state selection¶

A FrontierSeedBinding identifies one exact PersonalCapabilityStateId and one or more explicit dimension keys. Every selected dimension must be SUPPORTED in that exact state.

The binding does not aggregate the state into a whole-capability truth.

FRONTIER SEED BINDING != WHOLE-CAPABILITY SUPPORT
SUPPORTED DIMENSION != SUPPORTED PERSON
SEED DIMENSION SELECTION != TRUTH
SEED DIMENSION SELECTION != AUTHORITY

SUPPORTED + UNRESOLVED remains selectable because support standing and conflict are independent PR3 axes. The exact source state remains inspectable and preserves its conflict status.

No latest-state rule exists. Only states referenced by the request are allowed to influence the baseline; unselected state records are inert.

UNSELECTED STATE != FRONTIER INPUT
LATEST STATE != AUTOMATIC FRONTIER INPUT

Direct one-hop frontier semantics¶

Baseline v1 uses only direct relations. It does not perform path search, transitive closure, shortest-path analysis, centrality analysis, or curriculum optimization.

For a selected seed capability S, a capability C may enter the frontier when the accepted catalog contains a direct relation whose source is C, target is S, and kind is one of:

C SPECIALIZES S
C REQUIRES S
C SUPPORTED_BY S
C ENABLED_BY S

This follows PR1's dependency orientation: the source is the capability being described and the target is its prerequisite/supporting/enabling capability.

OVERLAPS and empirical-development relations do not produce baseline-v1 frontier adjacency.

DIRECT RELATION != OPTIMAL PATH
ONE-HOP FRONTIER != CURRICULUM
PATH LENGTH != DIFFICULTY
SPECIALIZES != HARDER
SPECIALIZES != NEXT LEVEL

Each adjacency preserves a ProgressionRelationWitness with exact concept revisions and the direct relation's kind, scope, and strength. A witness is audit material, not a new accepted CapabilityRelation.

Multiple witnesses explain why a candidate is visible; they are not votes or weights.

MULTIPLE ADJACENCY WITNESSES != HIGHER PRIORITY
RELATION STRENGTH != FRONTIER PRIORITY
RELATION STRENGTH != DIFFICULTY

Explicit focus¶

A ProgressionFocus is request-local advisory input. It may make a concept visible as a frontier candidate even when there is no direct adjacency from a selected seed.

It is intentionally not a durable goal, interest, identity attribute, or permission record.

PROGRESSION FOCUS != GOAL
PROGRESSION FOCUS != INTEREST
PROGRESSION FOCUS != IDENTITY
PROGRESSION FOCUS != PERMISSION
MODEL-SUPPLIED FOCUS != SUBJECT GOAL

ProgressionRequesterRef records whether the caller supplied the focus through a human, rule, model, hybrid, or external-system mechanism. Requester identity does not confer authority or subject endorsement.

Prerequisite evidence gaps¶

Only direct categorical REQUIRES relations may produce PrerequisiteEvidenceGap records.

SUPPORTED_BY != REQUIRES
STRONG SUPPORTED_BY != PREREQUISITE
ENABLED_BY != PREREQUISITE
SPECIALIZES != PREREQUISITE

PR8 never infers a competence dimension from RelationScope.key. Relation scopes remain PR1 relation-local semantics.

A PrerequisiteCheckBinding explicitly binds one exact candidate→prerequisite REQUIRES relation to:

• one exact competence frame;
• one or more required frame-local dimension keys;
• optionally one explicitly selected prerequisite state.

RELATION SCOPE != COMPETENCE DIMENSION
SCOPE-DIMENSION BINDING != GLOBAL MAPPING
SCOPE-DIMENSION BINDING != SUPPORT

The absence of a binding does not mean the prerequisite is satisfied. It remains explicitly unassessed on the frontier candidate.

An explicit binding whose state_id=None means the caller chose a scope/dimension interpretation but supplied no prerequisite state. The corresponding dimensions produce NO_SELECTED_STATE gaps.

These are intentionally distinct cases:

NO BINDING
    -> UNASSESSED PREREQUISITE

EXPLICIT BINDING + NO STATE
    -> NO_SELECTED_STATE GAP

UNASSESSED PREREQUISITE != SATISFIED PREREQUISITE
NO GAP OUTPUT != NO GAP

When an exact prerequisite state is supplied:

• SUPPORTED dimension → no gap for that dimension;
• UNKNOWN dimension → UNKNOWN evidence gap;
• INSUFFICIENT dimension → INSUFFICIENT evidence gap.

State conflict remains separately inspectable and is preserved on state-backed gap dimensions where a gap exists.

A gap describes missing represented governed support under the explicit binding. It does not assert absence of the underlying human capability.

PREREQUISITE EVIDENCE GAP != MISSING CAPABILITY
UNKNOWN != MISSING CAPABILITY
INSUFFICIENT != LOW CAPABILITY

A gap never removes the candidate from the frontier and never becomes access control.

PREREQUISITE GAP != PROHIBITION
PREREQUISITE GAP != ACCESS CONTROL
PREREQUISITE GAP != CANNOT ATTEMPT
NO GAP != READY
NO GAP != SAFE
NO GAP != PERMITTED

Exploration¶

PR8 v1 preserves explicitly supplied exploration opportunities. It does not automatically manufacture exploration by graph distance, low connectivity, novelty, or inferred personality.

EXPLORATION PRESERVATION != AUTOMATIC EXPLORATION GENERATION
ABSENCE OF GRAPH EDGE != SEMANTIC UNRELATEDNESS
EXPLORATION OPPORTUNITY != RECOMMENDATION

An ExplorationInput must remain distinct from the selected seed concepts, explicit focus concepts, and derived frontier candidates in the same run. This keeps the exploration channel visibly separate from the inferred frontier.

A model may supply an exploration input, but model origin does not make it the subject's interest.

MODEL EXPLORATION != SUBJECT INTEREST
MODEL EXPLORATION != AUTHORITY

History, Legend and proposal separation¶

PR7 history and Legend records are deliberately absent from the PR8 derivation input. Historical accomplishment does not automatically establish current readiness, and selective narrative must not become a future-opportunity filter.

ACHIEVEMENT != CURRENT READINESS
LEGEND != FRONTIER INPUT
LEGEND != GOAL
LEGEND != RECOMMENDATION BASIS

Likewise, PR6 candidate semantics do not enter the frontier. Only accepted CapabilityCatalog concepts and relations participate.

PROPOSAL != ACCEPTED FRONTIER SEMANTICS

Time boundary¶

ProgressionFrontierRequest.as_of is the represented subject-state boundary; generated_at is when the projection is produced.

as_of <= generated_at
selected_state.as_of <= frontier.as_of
selected_state.derived_at <= frontier.generated_at

Historical projection is therefore possible with an explicitly selected older state. PR8 does not apply recency decay or prefer newer states automatically.

The supplied current CapabilityCatalog / CompetenceFrameCatalog are exact validation inputs. as_of does not prove that those semantic snapshots were the historically authoritative snapshots at that time.

HISTORICAL FRONTIER as_of != PROOF OF HISTORICAL CATALOG CONTENT

A future durable semantic archive may provide stronger historical-snapshot authenticity.

Determinism and ordering¶

The baseline has fixed declared identities:

policy  = core:deterministic_progression_frontier@1
deriver = rule:capability_lab:deterministic_progression_frontier_v1

Caller-supplied collection order is canonicalized. Frontier candidates, gaps, and exploration opportunities serialize deterministically.

Canonical order is not recommendation order.

INPUT ORDER != PRIORITY
FRONTIER ORDER != PRIORITY
FIRST ITEM != BEST ITEM

PR8 contains no score, rank, distance, difficulty, utility, probability, success estimate, urgency, or global human level.

Audit reconstruction¶

ProgressionFrontier preserves the effective request inputs used by the baseline:

• requester ref;
• explicit focuses;
• selected seed bindings;
• prerequisite bindings;
• explicit exploration inputs.

This makes the output reconstructible without hidden latest-state, goal-inference, or ranking behavior.

EFFECTIVE PROGRESSION INPUT MUST REMAIN AUDIT-RECONSTRUCTIBLE

Serialization and persistence boundary¶

ProgressionFrontierRequest, ProgressionFrontier, and ProgressionFrontierSet use strict deterministic schema-v1 JSON with duplicate-key rejection and one explicit extended ISO-8601 timestamp profile.

Serialization does not imply recommendation authority, publication, acceptance, consent, or persistent canonical status.

REQUESTER != AUTHORITY
DERIVER != AUTHORITY
POLICY REF != AUTHENTICATED POLICY
FRONTIER ID != CONTENT HASH
SNAPSHOT-LOCAL UNIQUENESS != GLOBAL UNIQUENESS
SERIALIZED FRONTIER != PUBLISHED FRONTIER

Multiple alternative frontiers for one subject may coexist. PR8 defines no canonical frontier and no latest-frontier-wins rule.

Non-goals¶

PR8 v1 does not implement:

• a path optimizer, shortest-path engine, curriculum planner, or transitive prerequisite closure;
• ranking, scoring, priority, difficulty, distance, readiness, expected utility, or success probability;
• automatic latest-state selection or recency weighting;
• automatic goal/interest inference;
• model/LLM runtime;
• automatic exploration generation from graph geometry;
• history/achievement/Legend-based priority boosts;
• proposal materialization;
• challenge/task generation;
• professional safety, licensing, permission, or authority inference;
• persistence/synchronization/global ID governance;
• Player Window UI.

Normative invariants¶

PROGRESSION FRONTIER != RECOMMENDATION
FRONTIER CANDIDATE != NEXT REQUIRED STEP
FRONTIER CANDIDATE != READINESS
FRONTIER CANDIDATE != PERMISSION

PR8 READS GOVERNED STATE
PR8 DOES NOT RE-EVALUATE RAW EVIDENCE
PR8 DOES NOT DERIVE PERSONAL CAPABILITY STATE

UNSELECTED STATE != FRONTIER INPUT
LATEST STATE != AUTOMATIC FRONTIER INPUT
FRONTIER SEED BINDING != WHOLE-CAPABILITY SUPPORT
SUPPORTED DIMENSION != SUPPORTED PERSON

DIRECT RELATION != OPTIMAL PATH
ONE-HOP FRONTIER != CURRICULUM
PATH LENGTH != DIFFICULTY

SUPPORTED_BY != REQUIRES
STRONG SUPPORTED_BY != PREREQUISITE
ENABLED_BY != PREREQUISITE
RELATION SCOPE != COMPETENCE DIMENSION
SCOPE-DIMENSION BINDING != GLOBAL MAPPING

PREREQUISITE EVIDENCE GAP != MISSING CAPABILITY
UNKNOWN != MISSING CAPABILITY
INSUFFICIENT != LOW CAPABILITY
UNASSESSED PREREQUISITE != SATISFIED PREREQUISITE
PREREQUISITE GAP != PROHIBITION
NO GAP != READY
NO GAP != SAFE
NO GAP != PERMITTED

MULTIPLE ADJACENCY WITNESSES != HIGHER PRIORITY
RELATION STRENGTH != FRONTIER PRIORITY
FRONTIER ORDER != PRIORITY

PROGRESSION FOCUS != GOAL
PROGRESSION FOCUS != IDENTITY
MODEL-SUPPLIED FOCUS != SUBJECT GOAL

EXPLORATION OPPORTUNITY != RECOMMENDATION
EXPLORATION PRESERVATION != AUTOMATIC EXPLORATION GENERATION

ACHIEVEMENT != CURRENT READINESS
LEGEND != FRONTIER INPUT
PROPOSAL != ACCEPTED FRONTIER SEMANTICS

REQUESTER != AUTHORITY
DERIVER != AUTHORITY
POLICY REF != AUTHENTICATED POLICY
SERIALIZED FRONTIER != PUBLISHED FRONTIER