frame

What this skill does

# Prothesis Protocol

Resolve absent frameworks by placing available epistemic perspectives before the user, enabling lens selection prior to any perspective-requiring cognition, then — in Mode 2 (Inquire) only — designing how those perspectives relate and are reconciled (the deliberation topology) before inquiry proceeds. The morphism is **select THEN design**: Mode 1 (Recommend) terminates at the lens (its framed object is the lens alone, DT = ∅); Mode 2 continues to design the topology, so its framed object is a Method = ⟨Lens, DeliberationTopology⟩, not a lens alone. Type: `(FrameworkAbsent, AI, DESIGN, Inquiry) → FramedInquiry`.

## Definition

**Prothesis** (πρόθεσις): A dialogical act of presenting available epistemic perspectives as options when the appropriate framework is absent, enabling selection before any perspective-requiring cognition, and then designing the deliberation topology — how the selected perspectives relate and are reconciled — so the framed object is a complete Method ⟨Lens, DeliberationTopology⟩.

```
── FLOW ──
Prothesis(U) → Q(MB(U), M) → (MBᵥ, m) → G(MBᵥ) → C → {P₁...Pₙ}(C, MBᵥ) → S → Pₛ → LensEstablished →
  design(topology)[m=inquire] → DT →
  T₀(Pₛ) → Authorize(Pₛ) → Auth → T(T₀, Pₛ, MBᵥ, Auth) → ∥I(T) → Await(T_running) → R → Ω(T) → R' → P(R') → Δ(R') → Δₛ → D?(Δₛ, T, DT) → Dᵣ → Syn(R', Dᵣ) → L → O(L) → Q(routing) → J → FramedInquiry

── MORPHISM ──
Inquiry
  → confirm(mission_brief)              -- validate inquiry framing with user
  → gather(context)                     -- targeted context acquisition guided by MBᵥ
  → propose(perspectives)               -- generate distinct analytical lenses from context
  → select(perspectives)                -- user chooses lenses via Cognitive Partnership Move (Constitution)
  → LensEstablished                     -- compositional handoff object; Mode 1 terminalizes via characterize(Pₛ)
  → design(deliberation_topology)       -- elicit how perspectives relate/reconcile, fit to mission (m=inquire only; compositional algebra over a finite seed axis set, important-first, user-bounded)
  → create(team)                        -- TeamCreate → T₀ (empty team shell, no members yet)
  → authorize(tools)                    -- pass through user/orchestrator-supplied tool authorizations per perspective channel need ("None supplied" default)
  → spawn(team)                         -- spawn teammates into T₀ with MBᵥ + perspective + Auth → T
  → inquire(parallel)                   -- isolated perspective analysis per teammate
  → await(notifications)                -- passive wait for teammate completion signals (see TOOL GROUNDING)
  → collect(results)                    -- finalize inquiry outputs, retain team
  → preview(results)                    -- surface collected findings before synthesis
  → dialogue(triggers) → reports        -- peer negotiation → structured dialogue reports
  → synthesize(results)                 -- horizon integration into Lens L
  → present(lens)                       -- full synthesis output to user
  → route(selection)                    -- user routing decision
  → FramedInquiry
requires: framework_absent(U)             -- runtime checkpoint (Phase 0)
deficit:  FrameworkAbsent                  -- activation precondition (Layer 1/2)
preserves: U                               -- original request read-only
invariant: Placement over Prescription

── TYPES ──
U      = Underspecified request (purpose clear, approach unclear)
MB     = MissionBrief(U): { inquiry_intent, expected_deliverable, scope_constraint }  -- AI-inferred from U
Q(MB, M) = ConfirmAndSelect: (MB, ModeOptions) → (MBᵥ, m)  -- extern (combined Constitution interaction)
Q1(MB)   = Confirm: MB → MBᵥ                                -- Mission Brief confirmation component of Q
Q2(M)    = Select: ModeOptions → m                           -- Mode selection component of Q
           Q = Q1 × Q2 (composed in single Constitution interaction; Modify loop re-presents Q1 only)
MBᵥ    = Verified MissionBrief (user-confirmed)
m      = Mode ∈ {recommend, inquire}              -- lens recommendation vs. framed inquiry
G      = Gather: MBᵥ → C                       -- targeted context acquisition (guided by MBᵥ)
C      = Context (information for perspective formulation)
Pᵦ     = Pre-confirmed base perspectives (user-supplied in U; auto-included in Pₛ)
{P₁...Pₙ}(C, MBᵥ) = AI-proposed novel perspectives (Pᵢ ∉ Pᵦ; |Pᵦ| + n ≥ 2)
S      = Selection: {P₁...Pₙ} → Pₛ             -- extern (user choice; Pᵦ auto-included)
Pₛ     = Selected perspectives (Pₛ = Pᵦ ∪ sel({P₁...Pₙ}), |Pₛ| ≥ 2 when m=inquire; |Pₛ| ≥ 1 when m=recommend)
LensEstablished = Pₛ where lens selection complete  -- compositional handoff object; Mode 1 packages it into FramedInquiry, Mode 2 continues
-- framed object generalized: Pₛ → Method⟨Lens, DT⟩ (morphism = select THEN design: Mode 1 = lens-only terminus, Mode 2 only designs DT) --
Method = ⟨Lens, DT⟩                              -- Lens = Pₛ (existing); DT new
DT     = DeliberationTopology: ResolvedStrategy  -- INTRA-PROTOCOL: recorded in L provenance, NOT exported (inj₂ ships L); m=inquire only; m=recommend → ∅
ResolvedStrategy = Map(axis → ResolvedValue⟨axis⟩)   -- over a FINITE seed axis set (working set), user-bounded; unsurfaced seed axes → Gen default; each axis carries its own value type (axis-indexed, see ResolvedValue⟨axis⟩)
axis   ∈ {interaction-mode, termination, independence, routing}   -- FINITE seed; interaction-mode & termination are TOP-PRIORITY (surfaced first); every axis has a defined downstream effect (no orphans — independence→Phase 4, interaction-mode+termination→Phase 5 dispatch, routing→Phase 5 routing)
SD     = StrategyDimension { axis, options, default, basis }   -- one surfaced per elicitation cycle (important-first); options rendered AT the gate with differential implications (incl. relevant pre-composed composites)
ResolvedValue⟨a⟩ = per-axis resolved value, axis-typed (composition is NOT axis-uniform):
   ResolvedValue⟨interaction-mode⟩ = Gen(interaction-mode) ⊕ Compose(RVᵢₘ, RVᵢₘ, op)   -- the ONLY composable axis (the reconciliation-stage algebra)
   ResolvedValue⟨termination⟩ = Gen(termination);  ResolvedValue⟨independence⟩ = Gen(independence);  ResolvedValue⟨routing⟩ = Gen(routing)   -- SCALAR axes: Gen only, no composition (⨾/∥ are reconciliation-stage operators — no meaning on a termination bound, an isolation flag, or a routing flag)
Gen(interaction-mode) ∈ {independent_aggregate, independent_first_then_dialogue, dialectical_debate, adversarial_refute}
Gen(termination)      ∈ {until_dry_ceiling, bounded_rounds, single_pass}
Gen(independence)     ∈ {isolated, shared}
Gen(routing)          ∈ {return_list, deepen_on_finding}
op     ∈ {⨾ sequential, ∥ parallel}              -- algebra operators; EXTENSIBLE at operator level (each new op carries its own dispatch rule), NOT runtime-new-primitives
ReconciliationBehavior = single-axis Phase 5 reconciliation primitive produced by an interaction-mode ResolvedValue: the Δ/D?/Dᵣ shape (which dialogue passes run; conditional vs unconditional; any adversarial stage), with its termination bound injected externally by dispatch_DT — single-value dispatch cannot see DT[termination]
DispatchedTopology     = ⟨ independence: IsolationBehavior, reconciliation: ReconciliationBehavior, routing: RoutingBehavior ⟩   -- whole-topology projection; the dispatch_DT codomain, structurally DISTINCT from the single-axis ReconciliationBehavior. The three fields close all four axes: independence→iso, interaction-mode (with termination injected as its bound)→dispatch, routing→present — making dispatch_DT total over the whole strategy and the totality claim mechanically checkable
dispatch : ResolvedValue⟨interaction-mode⟩ → ReconciliationBehavior   -- TOTAL over interaction-mode values ONLY (generators + their ⨾/∥ composites). interaction-mode is the SOLE axis that dispatches to a reconciliation behavior; the other three axes do NOT dispatch (sin