settled-reach/docs/workshops/v01-content-scoping/round2-tracking.md

Round 2 Tracking — Qatux (Documenter)

v0.1 Content Scoping Workshop — CLOSING ROUND

Status: All 7 responses received (Gestalt, Paula, Tyre, Dudley, Stig, Mellanie, SI). Workshop complete. Date: 2026-02-12

---

1. Lead Decisions (Round 2 Brief)

Four lead decisions were issued at the start of Round 2, resolving the major disagreements from Round 1:

ID	Decision	Round 1 Disagreement Resolved
D-NPC	Gestalt's mechanical mapping is canonical for v0.1. Paula's emotional/narrative layer becomes v0.2 annotations.	D-02 (10 of 17 NPC mapping disagreements)
D-FORMAT	YAML for content files. RON converter available as optional build-time tool.	D-01 (YAML vs RON)
D-INTERACT	Single context-sensitive action for v0.1. Multi-verb architecture modeled underneath from day one.	D-04 (Examine vs Talk: separate or combined)
D-PAUSE	Spacebar always pauses (0% speed). UI overlays (knowledge panel, dialogue) trigger 50% sim speed. 3-state system.	D-05 (Knowledge panel pause behavior)

Additionally resolved by lead direction:

• D-03 (EntityKnowledge keys): 16 total keys. 4 new generic role-perspective keys. trust_read merged into existing trust_level. leverage_held absorbs secret_held. secret_confidence renamed to leverage_confidence. (See Section 5.)
• R2-06 (PC voice registers): Assigned to Mellanie as Round 2 task.
• R2-07 (Triangle fork cascade): Self-contained forks confirmed.
• R2-09 (ObserverSnapshot v3): Assigned to Tyre as Round 2 task.

Closing round additions (post-Round 2 responses):

• Q-012 resolved: 16 EntityKnowledge keys finalized (see Section 5).
• Q-013 resolved: Server-authoritative pause. Client is a dumb renderer. NEW-12 (client pause state machine) killed.
• Q-016 resolved: actions[] → verbs[] rename confirmed across server code, client code, YAML content, wiki docs, decision files. Folded into NEW-7.
• Process rule: No ticket merges across domain teams. Shared epics are fine; each team gets its own ticket with its own definition of done.

---

2. Round 1 Disagreement Resolutions

D-02 Resolved: NPC Pattern/Motivation Mapping

Gestalt reviewed all 10 disagreements from Round 1 and changed his position on 6 of 10, adopting Paula's reads. Both agents converged on a single canonical mapping.

NPC	Round 1 Gestalt	Round 1 Paula	Round 2 Final	Who Prevailed	Gestalt's Rationale for Change
Maret (pattern)	SYSTEM	WITNESS	WITNESS	Paula	Story function > job description
Maret (motivation)	WITNESS	SKEPTIC	SKEPTIC	Paula	WITNESS+SKEPTIC creates compositional tension
Resha (motivation)	SKEPTIC	CIVILIAN	CIVILIAN	Paula	Innocent disruption > deliberate questioning
Harek (pattern)	GHOST	ANCHOR	ANCHOR	Paula	Harek is physically present; contentment, not absent power
Drin (pattern)	REMNANT	SYSTEM	SYSTEM	Paula	Station too young for REMNANT; SYSTEM+TURNCOAT = institutional rot
Sabel (pattern)	THRESHOLD	GHOST	GHOST	Paula	Off-station = absent power, not gatekeeper
Olin (motivation)	SKEPTIC (dual)	CIVILIAN	CIVILIAN	Paula	Dropped dual designation; 1 pattern + 1 motivation per NPC
Voss (motivation)	OPERATOR	HANDLER	OPERATOR	Gestalt	Voss RUNS the schedule; middle management operating machinery
Devra (motivation)	HANDLER	OPERATOR	HANDLER	Gestalt	Devra conveys Nils's strategy; pressure from above
Sess (pattern)	ANCHOR	NOBODY	ANCHOR	Gestalt	NOBODY deferred to v0.2; ANCHOR at T3 = contentment as texture

Paula's response: Accepted the versioned approach. Annotated all 10 disagreement NPCs with v0.2 emotional layer descriptions. Each annotation specifies what content the v0.2 motivation/pattern unlocks (~30-50 lines per NPC).

D-01 Resolved: YAML Format

Dudley accepts YAML. One-line change: serde_yaml::from_str() replaces ron::from_str(). Multi-line YAML blocks (|) improve author experience. RON crate retained for potential future use.

D-04 Resolved: Examine vs Talk

Converged on: conceptually separate, v0.1 ships single action. The server models both verbs internally; priority resolution determines which surfaces as the [E] prompt. Post-contradiction, Examine can become primary (detective investigating a flagged NPC). The v0.2 client reads the full action list — no server changes needed.

D-05 Resolved: Pause System

3-state system adopted by all respondents. Gestalt, Stig, and Dudley each independently designed compatible state machines.

---

3. Definitive v0.1 NPC Mapping

The canonical mapping, produced by Gestalt with 6 changes from Paula's input:

#	Name	Tier	Pattern	Motivation	Composition Read
1	Kael Davan	T1	FRIEND	OPERATOR	Trust anchor running ring cargo
2	Sera Venn	T1	FRIEND	WITNESS	Trust anchor carrying unreported evidence
3	Naia Tamm	T1*	MIRROR	CIVILIAN	Honest transparency, no secrets
4	Voss	T2	SYSTEM	OPERATOR	Institutional authority operating the schedule
5	Lera Sessik	T2	ANCHOR	OPERATOR	Contentment embodied, running a bar on willful ignorance
6	Torek Lintar	T2	CATALYST	CIVILIAN	Spending disrupts equilibrium
7	Devra	T2	THRESHOLD	HANDLER	Ring gatekeeper, manages operations from above
8	Maret Korr	T2	WITNESS	SKEPTIC	Accidentally observed discrepancies, doubts the numbers
9	Resha	T2	CATALYST	CIVILIAN	New hire forcing defensiveness with innocent questions
10	Drin	T2	SYSTEM	TURNCOAT	Inspector whose compromise corrupts the institution
11	Renn	T2	THRESHOLD	OPERATOR	Courier controlling cargo flow
12	Pell	T2	CATALYST	TURNCOAT	Wavering member forcing potential crisis
13	Harek	T2	ANCHOR	HANDLER	Complacent contentment whose comfort creates a tool
14	Sess	T3	ANCHOR	CIVILIAN	Quiet contentment behind the bar
15	Olin	T3	CATALYST	CIVILIAN	New arrival forcing evaluation
16	Sabel	T3	GHOST	OPERATOR	Off-station absent power running the supply chain
17	Tav	T3	SYSTEM	CIVILIAN	Ring lookout following procedure

Off-stage: Nils Davan — GHOST + HANDLER

Distribution: 8 of 9 patterns present (REMNANT correctly absent; NOBODY deferred to v0.2). All 6 motivations present. CIVILIAN at 6/17 (35%) matches D-029 noise floor (~30%).

---

4. Key Architecture Convergences

4.1 Multi-Verb Interaction Architecture

Tyre, Dudley, and Stig independently produced compatible designs:

Agent	Contribution	Compatible?
Tyre	AvailableActions per entity with priority-sorted ActionOption list; ActionKind enum; InteractPrimary vs InteractWith PlayerAction variants	Canonical design
Dudley	EntityInteractions with AvailableAction list; server-side rank_actions() priority system; PlayerAction::InteractWith for v0.2	Compatible — naming differs, structure identical
Stig	InteractionOptions with primary/secondary fields; v0.2 extension reads actions[] array; world-space prompt renderer	Client-side compatible

Minor naming divergence: Tyre uses AvailableActions/ActionOption, Dudley uses EntityInteractions/AvailableAction. These are the same concept.

Lead confirmation (closing round): actions[] → verbs[] rename confirmed. Scope: server code, client code, YAML content files, wiki docs, decision files. Folded into NEW-7 ticket description.

Key design principles (unanimous):

• Server computes N available verbs per entity every tick
• v0.1 client shows only priority-0 verb (single [E] prompt)
• v0.2 client renders full list — no server changes needed
• Context-sensitive priority (post-contradiction: Examine rises above Talk)
• InteractionOptions/AvailableActions embedded in ObserverSnapshot, not separate message
• Server is authoritative; client never queries content files

4.2 Pause System (3-State)

State	Speed	Trigger	All 3 Agree?
Normal	100% (1.0)	Default	Yes
Overlay	50% (0.5)	Knowledge panel, dialogue	Yes
Paused	0% (0.0)	Spacebar (always)	Yes

Gestalt: Multiplicative toggle model (effective_speed = if pause_override { 0.0 } else { base_speed * ui_modifier }). Multiple overlays don't stack (50% stays 50%). v0.2 extension: panel + dialogue could = 25%.

Stig: GDScript PauseManager with SimState enum, _pre_pause_state tracking for resume, overlay count tracking. Spacebar from OVERLAY_ACTIVE goes to PAUSED (not RUNNING). Questions for Dudley: does server need pause signals, or does client control?

Dudley: SimSpeed enum resource with speed_stack (priority-based). SimSpeed::Overlay ticks every 2nd frame. Server is authoritative over tick advancement. sim_speed: f32 field in ObserverSnapshot tells client the current rate.

Resolved (closing round): Server-authoritative pause. Tick rate is a server simulation concern. Client is a dumb renderer — it renders ObserverSnapshots at whatever rate the server sends them. Client sends a pause IPC command to the server; that's it. Stig's PauseManager GDScript is reduced to sending IPC commands and reading sim_speed from the snapshot for visual feedback. NEW-12 (client pause state machine) killed — client doesn't manage time.

4.3 ObserverSnapshot v3

Tyre and Dudley independently produced v3 definitions. They are structurally identical:

Field	Tyre's Proposal	Dudley's Proposal	Status
sim_speed: f32	Yes (default 1.0)	Yes	Agreed
nearby_interactions	Vec<NearbyInteraction>	Vec<EntityInteractions>	Agreed (naming differs)
active_dialogue	Option<ActiveDialogue>	Option<ActiveDialogue>	Agreed
monologue	Vec<MonologueDisplay>	Vec<MonologueDisplay>	Agreed
overheard	Vec<OverheardFragment>	Vec<OverheardDisplay>	Agreed (naming differs)
knowledge_updates	Vec<KnowledgeUpdate>	Not in Dudley's R2 snapshot	Tyre adds delta updates
examine_result	Option<ExamineResult>	Not in Dudley's R2 snapshot	Tyre adds
ticker_headlines	Vec<TickerHeadline>	Not in Dudley's R2 snapshot	Tyre adds

Wire size estimate (Tyre): v3 adds ~50-1500 bytes per tick on top of v2's 2-8 KB. Total: ~2-10 KB/tick at 10 tps = 20-100 KB/s. Well within IPC budget.

Backward compatibility: MessagePack serializes by field name. New fields use #[serde(default)]. v2 client reading v3 snapshot ignores unknown fields. v3 client reading v2 snapshot gets empty defaults.

4.4 Content Directory + Schema

Tyre produced the definitive directory tree and JSON Schema definitions:

Directory structure: content/content.yaml (manifest) + content/schema/ (JSON Schema files) + content/global/ (facts, factions, enums, entity-schema) + content/districts/sova-transit/ (npcs/, locations/, triangles/, dialogue/, monologue/, routines/).

Schemas defined:

• npc-profile.schema.json — Full NPC profile with tier-conditional fields, pattern/motivation enums, axes, tells, access tiers, trust levels, friend_arc, triangle membership
• dialogue-pool.schema.json — Tagged dialogue lines per D-028/D-035. Role + location based, with access/trust/situation gating and knowledge_grants
• monologue-pool.schema.json — Character-partitioned (D-032). Trigger + prerequisite (AND-only) + priority + cooldown. 160 char max per Stig

Validation pipeline (3-tier):

1. Authoring time: IDE with JSON Schema autocomplete
2. Build time (make content-validate): JSON Schema + cross-reference validation
3. Load time: serde_yaml deserialization + semantic validation (StableId assignment, relationship wiring)

4.5 YAML + RON Converter

Aspect	Tyre	Dudley	Status
Content format	YAML	YAML (accepted)	Agreed
RON converter location	tooling/content-tools/	tooling/content-converter/	Minor naming difference
When converter runs	Build-time optional	Build-time optional, low priority	Agreed
Development loading	YAML direct	YAML direct	Agreed
Production loading	RON cache preferred, YAML fallback	YAML (RON is insurance)	Agreed — YAML in v0.1, RON as future optimization

Tyre provides dual-format loader (~20 lines) with load_with_cache() fallback pattern. content/.cache/ is gitignored.

---

5. EntityKnowledge Keys — RESOLVED: 16 Keys

Both Gestalt and Paula independently arrived at the same conclusion: generic perspective keys, not per-archetype keys. This resolves D-03 from Round 1.

Convergence (Round 2 Responses)

Point	Gestalt	Paula	Converged?
Per-archetype keys are unsustainable	Yes (32 archetype-specific + 14 shared = 46 total)	Yes (same conclusion)	Yes
Generic keys work for all 8 archetypes	Yes	Yes	Yes
Interpretation lives in content, not schema	Yes (monologue pools provide character-specific meaning)	Yes (archetype brief + voice register + moral arc)	Yes
social_debt deferred to v0.2	Yes	Accepts	Yes
loyalty_conflict deferred to v0.2	Yes (inferable from behavior_flags + reliability)	Accepts	Yes

Final Resolution (Closing Round Lead Decision)

16 total EntityKnowledge keys. 12 existing shared keys + 4 new generic role-perspective keys. Key merges and renames applied:

4 New Role-Perspective Keys (observer→target, all archetypes)

Key	Value Enum	What It Stores
risk_assessment	none, low, moderate, high, critical	How this NPC threatens the observer's position (exposure, obstruction, liability, competition — archetype-dependent)
loyalty_assessment	solid, dependable, uncertain, wavering, hostile	How predictably this NPC serves the observer's interests
position_integrity	solid, thin, cracking, blown, N/A	Status of active deception involving this NPC (cover, reputation, facade)
moral_weight	innocent, peripheral, complicit, compromised, willing	Observer's subjective moral assessment within their ethical framework

Key names adopt Paula's naming (risk_assessment over threat_to_position, position_integrity over facade_status). Value enums adopt Gestalt's archetype-neutral vocabulary.

Key Merges and Renames

Action	Old	New	Rationale
Merge	trust_read	trust_level (existing)	"Read is the action, level is the end state." Same key, per-character interpretation.
Absorb	secret_held	leverage_held (existing, renamed)	"Secret is a subtype of leverage." Value space expanded: secrets, debts, promises, obligations, compromising positions.
Rename	secret_confidence	leverage_confidence	Follows secret_held → leverage_held rename.
Defer	social_debt	v0.2	30-minute session doesn't develop favor economy.
Defer	loyalty_conflict	v0.2	Inferable from behavior_flags + loyalty_assessment.
Not added	tell_observed, contradiction_flagged	—	Tracked in behavior_flags (existing). No 7th key needed.

Final Key Inventory (16 total)

#	Key	Category	New/Existing
1	name	Identity	Existing
2	role	Identity	Existing
3	faction	Identity	Existing
4	species	Identity	Existing
5	location	Spatial	Existing
6	routine_pattern	Behavioral	Existing
7	trust_level	Relational	Existing (absorbs trust_read)
8	leverage_held	Relational	Existing (renamed from secret_held, expanded value space)
9	leverage_confidence	Relational	Existing (renamed from secret_confidence)
10	relationship_state	Relational	Existing
11	behavior_flags	Behavioral	Existing (tracks tells + contradictions)
12	known_associates	Social	Existing
13	risk_assessment	Role-perspective	NEW
14	loyalty_assessment	Role-perspective	NEW
15	position_integrity	Role-perspective	NEW
16	moral_weight	Role-perspective	NEW

This scales to all 8 archetypes without modification. Interpretation lives in content (monologue pools, voice registers, archetype briefs), not schema. The smuggler reads risk_assessment: "high" as exposure risk; the detective reads it as obstruction risk; the administrator reads it as political liability. Same key, same enum, different meaning.

---

6. FriendArc Phase Transitions (Dudley)

Dudley produced the definitive 4-phase state machine with explicit trigger code:

Phase Transition	Trigger	Content Dependency
Warmth → Trust	Interaction count (>=3) + time elapsed (>=600 ticks / 1 game-hour)	5-8 casual dialogue + 3-5 positive monologue
Trust → Doubt	Knowledge threshold — observer learns a doubt-triggering fact at KnowsOf+	5-8 deeper dialogue + 3-5 questioning monologue
Doubt → Conflict	Direct observation of contradiction (contradiction_discovered flag)	1 critical monologue line (wow moment #3) + 8-12 confrontation dialogue

Contradiction spatial staging: Server schedules NPC routine deviation (Kael in Corridor B-7 during shift transition). The check_contradiction_observation system checks: (1) FRIEND at contradiction location, (2) within time window, (3) bonded character has Direct confidence (LOS). All 5 state changes must happen atomically within one tick: knowledge graph updated, FriendArc phase advanced, RelationshipState changed (Friendly → PersonOfInterest), monologue triggered (urgent), entity color shifted (green → amber).

Content required per FRIEND NPC: ~35-46 authored lines minimum. Both FRIEND NPCs: ~70-92 lines. Aligns with D-034 estimate.

New component: InteractionLog — BTreeMap tracking (target_stable_id, tick) pairs. Supports Warmth→Trust transition and future NOBODY tier promotion.

---

7. PC Voice Registers (Mellanie)

Mellanie produced canonical voice register specifications for both playable characters:

Smuggler — Internal Voice Register

One-line summary: Watches people, not systems. Thinks in fragments. Worries about friends practically, physically, with the knot in the stomach.

Parameter	Value
Sentence length	Short. Fragments welcome. Rarely >2 sentences.
Vocabulary	Concrete, physical, sensory. Trade jargon without explanation.
Emotional register	Feeling-first, analysis second.
Relationship mode	First-name basis. People defined by connection, not role.
Observation priority	People > environment > objects.
Self-reference	Rare. Through physical sensation, not emotional vocabulary.
Humor	Dry, situational, understated. Dark when stressed.

5 voice anchors provided: Baseline warm, Operational, Concerned, Hurt, Atmospheric. 4 anti-patterns defined: Too clinical, institutional vocabulary, analytical distance, explaining known facts.

Detective — Internal Voice Register

One-line summary: Categorizes first, feels second. Thinks in complete sentences. Institutional training runs in the background.

Parameter	Value
Sentence length	Complete but not long. More structured than smuggler.
Vocabulary	Institutional layer over plain speech. "Noted," "flagged," "filed it."
Emotional register	Analysis-first. Personal breaks through as notable cracks.
Relationship mode	Surname-first for targets. First-name for allies. The shift IS a marker.
Observation priority	Patterns > people > environment. Counts things, tracks frequencies.
Self-reference	More common, always through professional lens.
Humor	Wry, observational. Appreciates irony at a distance.

5 voice anchors provided. 4 comparison examples (same trigger, two characters) demonstrating voice distinctness.

Display Constraints (Mellanie confirms Stig's R1 proposal)

Constraint	Value
Max length	~160 characters (hard ceiling)
Target range	80-120 characters (most lines), 40-80 (ideal)
Max visual lines	2
Word target	10-18 words (14 average)
Display time	4-6 seconds (length-adjusted)
Character tint	Smuggler: amber-white #f0e0c8, Detective: blue-white #c8d8f0

---

8. THE FRIEND Content Pack — Kael Davan (Mellanie)

Mellanie mapped the complete Kael Davan content pack structure:

Content Type	Lines	Phase Distribution
Kael dialogue	36	Terminal: 17, Bar: 13, Corridor: 6
Smuggler monologue about Kael	18	5 phases (Comfort 4, First Crack 4, Contradiction 3, Confrontation 3, Contaminated Trust 4)
Detective monologue about Kael	10	Identification 2, Behavioral 3, Suspicion 3, Informant 2
Tell observation lines	8	Lattice 2, Shortened 2, Looks-left 2, Spatial 2
Trust-gated lines	13	Insider 6, Close colleague 4, Secret hints 3
Unprompted lines	6	Early warmth 3, Late warnings 3
PACK TOTAL	91	Within D-034 range (70-100)

Phase distribution analysis:

• Phase 1 (Comfort): 21 lines (23%) — heavy on dialogue, light on monologue
• Phase 2 (First Crack): 13 lines (14%) — tells ramp up
• Phase 3 (Contradiction): 7 lines (8%) — fewest lines, highest stakes
• Phase 4 (Confrontation): 16 lines (18%) — dialogue-heavy
• Phase 5 (Contaminated Trust): 16 lines (18%) — voice discipline critical
• Cross-phase: 16 lines (18%)

Sample YAML lines provided for each phase demonstrating prerequisite progression. Format aligns with Tyre's monologue-pool.schema.json.

695 total validation (Mellanie clarifies): 695 is the authored floor. Does not include generation-expanded output (D-028: "write 10, generate 40" for dialogue). Does not include engine-generated content (routine greetings, time-of-day variations).

---

9. UI Specifications Confirmed (Stig)

Updated Bill of Materials

System	Scenes	Scripts	Shaders	Priority
Monologue renderer	1	1	0	P0
Interaction prompt	1	1	0	P0
Dialogue panel	1	1	0	P0
Entity color system	0	1	1	P0
Pause manager	0	1	0	P0
HUD bar	1	1	0	P1
Minimap	1	1	1	P1
Knowledge panel	1	1	0	P1
News ticker	1	1	0	P1
Fog rendering	0	0	1	P0
TOTAL	7	9	3

Changes from Round 1: Reduced from 9 scenes to 7. Added 1 shader (fog rendering). Pause manager added (P0, new in R2).

Monologue Display Parameters (Stig, confirmed by Mellanie)

Parameter	Value
Max characters	160
Max lines	2
Base duration	4.0s
Short duration (<80 chars)	3.0s
Urgent chime duration	5.0s
Fade in	0.3s
Fade out	0.5s
Interruption fade	0.15s
Cooldown between lines	2.0s minimum
Max queue depth	1 (drop lower-priority)

Priority ordering: observe_anomaly (1) > witness_interaction (2) > discover_evidence (3) > post_conversation (4) > hear_sound (5) > observe_npc (6) > enter_location (7) > return_visit (8) > time_idle (9).

Critical UI interaction: Monologue fires during dialogue (for wow moment #3). Monologue sits in world layer; dialogue sits in HUD layer. Two voices simultaneously — one external, one internal.

Interaction Prompt — Single-Action Extensible

v0.1: [E] Talk / [E] Examine — single keybind, context-sensitive verb. v0.2: [E] Talk [F] Observe — horizontal label row, same prompt position.

The verbs[] field (renamed from actions[] per D-060) exists in the protocol from day one. The v0.1 client ignores it. The v0.2 client reads it. No protocol change needed.

---

10. Ticket Analysis — Finalized (SI)

Lead Decision Impact on Tickets

Decision	Ticket Impact
D-NPC	NEW-1 simplified to "apply Gestalt's mapping." Paula's layer → v0.2 annotation.
D-FORMAT	Confirms C-series as YAML. Adds NEW-8 (RON converter).
D-INTERACT	NEW-7 reframed. Adds NEW-9 (server InteractionOptions) + NEW-10 (client extensible prompt).
D-PAUSE	Adds NEW-11 (server tick rate scaling). NEW-12 (client pause state machine) killed in closing round — client is a dumb renderer.

Final Ticket Count

Category	Count	Teams
Wiki fixes (A-series)	8	copy
Style guides & specs (B-series)	5	copy
Content directory & schema (C-series)	10	copy (1), server (8), ci (1)
Design specs (D-series)	2	server (1), copy (1)
Round 1 new tickets (NEW 1-7)	7	copy
Round 2 new tickets (NEW 8-13)	5	server (4), client (1)
Total new tickets	37	copy (21), server (13), client (1), ci (1)
Existing ticket updates	10	—

Delta from Round 1: +5 tickets (NEW-8 through NEW-13 from lead decisions, minus NEW-12 killed in closing round). Removed: D3 (killed), A5 (deferred), D1 (deferred), NEW-12 (killed — client doesn't manage time). Merged: A3 into #319, #310 into B4.

Sprint 3 Allocation

Copy team (15 tickets):

• P0: #261 (Dual Lens Guide, critical path root), C1 (Content directory design), A1 (canonical names), A2 (Naia Tamm rename), A7 (smuggler attributes)
• P1: B2 (MIRROR spec), B4 (smuggler voice + moral arc), NEW-1 (NPC mapping), NEW-5 (wow moment checklist), A10 (Nils stub), A11 (Triangle 1 fix)
• P2: B3 (PC-as-NPC), A8 (shortcode fix), NEW-2 (Drin expansion), NEW-7 (interaction verb spec)

Server team (4 additions): C2 (directory skeleton), C3 (schemas), NEW-9 (InteractionOptions), NEW-11 (tick rate scaling)

Client team (1 addition): NEW-10 (extensible prompt). NEW-12 killed — client sends pause IPC command; server manages tick rate.

Critical Path (Revised)

Two parallel chains:

Chain 1 — Content Authoring:

#261 (Dual Lens Guide) [P0, ROOT]
  -> B4 (Smuggler voice + moral arc)
    -> A6 (Smuggler FactId text) [Sprint 4]
    -> NEW-4 (Monologue pool ~200 lines) [Sprint 5]
  -> #297 (FRIEND: Kael) [Sprint 5]
  -> #298 (FRIEND: Sera) [Sprint 5]
  -> #190, #191, #192 (Content packs) [Sprint 5]

Chain 2 — Content Architecture:

C1 (Content dir design) [P0]
  -> C2 (skeleton) + C3 (schemas) [Sprint 3]
    -> C4-C8 (YAML population) [Sprint 4]
    -> NEW-3 (wiki->YAML conversion) [Sprint 4]
    -> #326 (YAML loader) [Sprint 4]
    -> C9 (validator) [Sprint 4]

Independent chains (Sprint 3):

• A10 (Nils stub) → A11 (Triangle 1 fix)
• NEW-9 (InteractionOptions, server) → NEW-10 (extensible prompt, client)
• NEW-11 (tick rate scaling, server) — no client counterpart (client reads sim_speed from snapshot)

Risks (SI)

1. #261 remains critical path root. Every content authoring ticket flows through it. Recommended: assign Paula, time-box 5 days.
2. Sprint 3 copy team has 15 tickets. 7 are small/parallelizable. Real load: #261, C1, B2, B4, A11.
3. Server Sprint 3 already heavy (10 existing + 4 new). C2/C3 blocked by C1 anyway.
4. Content conversion (NEW-3) is Sprint 4 bottleneck. Start with 3 Tier 1 NPCs as soon as C2 lands.

---

11. Wiki Fix Execution Plan (Paula)

Paula produced execution-ready content changes for each priority level:

Priority	Change	Files	Status	Effort
P0-1	Naia Tamm rename	16 files	Exact per-file changes specified	~2 hours
P0-2	Nils Davan stub	1 new + 3 cross-ref	Full draft ready	~30 min
P0-3	Drin Tier 2 expansion	1 rewrite + index	Full draft ready	~1 hour
P1-1	Canonical names	10 headers + index	Name table ready; 3 need Miri validation	~1 hour
P1-2	Triangle 1 escalation fork	index + 2 NPC notes	Self-contained version (v0.1)	~1 hour
P2-1	Smuggler-lens sections (17 NPCs)	17 profiles	Template ready; per-NPC content needed	~8-10 hours
P2-2	PC-as-NPC profiles	2 new files	Gap identified; not yet drafted	~3-4 hours
P2-3	FactId catalog smuggler entries	1 file	Gap identified; not yet drafted	~2 hours

P0-1 and P0-2 can start immediately. P0-3 blocked by A-01 formalization (consensus confirmed). P1-1 blocked by Miri validation (3 names). P1-2 depends on P0-2.

P1-2 (Triangle 1) — self-contained revision: Cross-triangle cascade removed per D-INTERACT. Path B no longer triggers Maret's discovery. Path C no longer makes detective investigation easier. Emotional weight preserved within Triangle 1.

---

12. Dependencies Matrix

Agent	Waiting On	From	Blocks	Status
Stig	IPC message format (ObserverSnapshot v3 struct)	Dudley/Tyre	All client UI	Open
Stig	Pause architecture decision	Dudley	Pause manager	Resolved — server-authoritative
Stig	Font selection, diegetic styling	Araminta	Knowledge panel visual	Open
Stig	First monologue content (5-10 lines)	Mellanie	Monologue renderer testing	Open
Dudley	ObserverSnapshot v3 sign-off	Tyre	Protocol implementation	Open
Dudley	FactId catalog (machine-readable)	Gestalt/Paula	Prerequisite evaluation	Open
Dudley	FRIEND content packs (phase-tagged)	Mellanie	FriendArc integration testing	Open
Dudley	Contradiction spatial data (location, time window)	Paula	Spatial staging system	Open
Mellanie	Content directory confirmed	Tyre	File structure for packs	Open
Mellanie	FactId catalog stabilized	Gestalt/Paula	Prerequisite placeholders	Open
Mellanie	EntityKnowledge key resolution	Lead	Prerequisite vocabulary	Resolved — 16 keys, 4 role-perspective
Paula	Miri validation (3 NPC surnames)	Miri	Canonical name application	Open

---

13. Decisions Ready to Formalize

From Lead's Round 2 Decisions

Proposed ID	Decision	Source
D-049	YAML is the content file format for v0.1. RON converter available as optional build-time tool in tooling/. Server loads YAML directly in development; RON cache is a production optimization.	D-FORMAT lead decision; Tyre design; Dudley acceptance
D-050	Gestalt's NPC pattern/motivation mapping is canonical for v0.1. Paula's emotional/narrative layer becomes v0.2 annotations. Version-toggle approach: same NPCs, different analytical depth per version.	D-NPC lead decision; Gestalt definitive table; Paula versioned layer
D-051	v0.1 ships single context-sensitive action; multi-verb architecture modeled underneath. Server computes N available actions per entity. v0.1 client shows priority-0 only. v0.2 client reads full list. No server changes for v0.2 multi-verb.	D-INTERACT lead decision; Tyre architecture; Dudley implementation; Stig client design
D-052	3-state pause system: Normal (100%), Overlay (50%), Paused (0%). Server-authoritative. Spacebar always pauses from any state. UI overlays (knowledge panel, dialogue) trigger 50% speed. Multiple overlays don't stack below 50%. Spacebar resumes to pre-pause state. Server controls tick rate; client is a dumb renderer. Client sends pause IPC command to server. Client reads sim_speed from ObserverSnapshot for visual feedback. NEW-12 (client pause state machine) killed.	D-PAUSE lead decision + closing round resolution; Gestalt mechanical design; Dudley server SimSpeed authoritative
D-053	Self-contained triangle forks for v0.1. Independent decision state machines, shared NPC state. No cross-triangle cascade. Players observe cross-triangle effects emergently through NPC behavior. v0.2 adds storyteller-managed cascade.	Gestalt confirmation; Tyre recommendation (R1); Paula revised Triangle 1
D-054	ObserverSnapshot v3 protocol. Extends v2 with: sim_speed, nearby_interactions, active_dialogue, monologue, overheard, knowledge_updates, examine_result, ticker_headlines. Backward compatible via serde defaults. Wire overhead ~50-1500 bytes/tick.	Tyre full definition; Dudley compatible definition
D-055	16 EntityKnowledge keys total. 4 new generic role-perspective keys: risk_assessment, loyalty_assessment, position_integrity, moral_weight. All observer→target. trust_read merged into existing trust_level. secret_held renamed to leverage_held with expanded value space (secrets, debts, promises, obligations, compromising positions). secret_confidence renamed to leverage_confidence. social_debt and loyalty_conflict deferred to v0.2. tell_observed and contradiction_flagged tracked in existing behavior_flags — no dedicated keys. Scales to all 8 archetypes without schema changes.	Gestalt generic architecture; Paula generic architecture; closing round lead resolution
D-056	PC voice registers defined. Smuggler: feeling-first, fragments, concrete/physical, first-name basis. Detective: analysis-first, complete sentences, institutional vocabulary, surname-first for targets. Both characters have 5 voice anchors, 4 anti-patterns, and 4 comparison examples.	Mellanie canonical specifications
D-057	Content directory structure: content/ with schema/, global/, districts/ top-level split. JSON Schema validation at build time. Dialogue by location + role (not NPC). Monologue partitioned by character. canonical_id format: npc:{slug}, loc:{district}:{location}.	Tyre finalized tree + schema definitions
D-058	THE FRIEND content pack template: Kael Davan. 91 lines across 5 arc phases + cross-phase. Pack validates tag taxonomy, load correctness, trigger appropriateness. Structure becomes template for all subsequent content packs.	Mellanie pack structure + sample YAML
D-059	Monologue display constraints. 160 char max, 2-line max, 4-6s display time, 2s cooldown, queue depth 1 (drop lower priority), 9-level priority ordering with observe_anomaly highest.	Stig constraints; Mellanie adoption
D-060	actions[] renamed to verbs[] across all surfaces. Scope: server code, client code, YAML content files, wiki docs, decision files. Folded into NEW-7 ticket description.	Closing round lead decision
D-061	No ticket merges across domain teams. Shared epics are fine, but each team gets its own ticket with its own definition of done.	Closing round lead decision (process rule)

From Round 1 (Carried Forward — Unchanged)

Proposed ID	Decision	Source
D-042	Drin promoted from Tier 3 to Tier 2	Round 1 consensus (A-01)
D-043	THE NOBODY mechanic deferred to v0.2; hidden data ships in v0.1	Round 1 consensus (A-02)
D-044	v0.1 interaction model: 7 interactions (Move, Look, Monologue, Examine Object, Examine NPC, Talk, Overhear)	Round 1 consensus (A-04)
D-045	v0.1 scope: add News ticker, PC-as-NPC, Time progression, Relationship state transitions to IN list	Round 1 consensus (A-14, A-15, A-16)
D-046	v0.1 scope: add Inventory, Stealth, Combat, Player-facing save/load, Lattice modification to explicit OUT list	Gestalt (Round 1)
D-047	v0.1 triangles: 3 active forks (T1, T2, T4), 2 passive tensions (T3, T5)	Round 1 consensus (A-07)
D-048	Client receives all text from server via state updates; client does not load or evaluate content files	Round 1 consensus (A-12)

---

14. Open Items

Resolved in Closing Round

#	Item	Resolution	Decision
Q-012	EntityKnowledge generic key count: 4 vs 6	16 total keys. 4 new role-perspective (risk_assessment, loyalty_assessment, position_integrity, moral_weight). trust_read → trust_level (merge). secret_held → leverage_held (absorb + expand). secret_confidence → leverage_confidence (rename). tell_observed/contradiction_flagged tracked in behavior_flags. No 7th key.	D-055
Q-013	Pause authority: client vs server	Server-authoritative. Client is a dumb renderer. Sends pause IPC command. Reads sim_speed from snapshot. NEW-12 killed.	D-052 (updated)
Q-016	actions[] → verbs[] scope	Confirmed across all surfaces: server code, client code, YAML content, wiki docs, decision files. Folded into NEW-7.	D-060

Still Open

#	Item	Owner	Status	Priority
Q-014	3 NPC surnames awaiting Miri validation: Drin Vassar, Sess Torin, Tav Lenn.	Miri	Blocked on world-building	Low
Q-015	Interaction struct naming: AvailableActions/ActionOption (Tyre) vs EntityInteractions/AvailableAction (Dudley). Trivial — needs one canonical choice.	Tyre/Dudley	Sprint 3 implementation	Low
Q-017	Mellanie's 695 authored items — validated by SI as scope input but not independently verified by other agents.	SI/Mellanie	Informational	Low

---

15. Workshop Summary

What This Workshop Produced

1. Canonical NPC mapping — 17 NPCs with pattern + motivation, through collaborative resolution (Gestalt changed 6 of 10 positions based on Paula's arguments)
2. Content architecture — Full directory tree, 3 JSON Schema definitions, 3-tier validation pipeline, dual-format loader
3. Interaction architecture — Multi-verb server model with single-action client filter, extensible without server changes
4. Pause system — 3-state model with unanimous design convergence across 3 agents
5. ObserverSnapshot v3 — Full protocol definition with backward compatibility, produced independently by 2 agents with structural agreement
6. PC voice registers — Canonical specifications for both characters with voice anchors, anti-patterns, and comparison examples
7. FRIEND content pack template — 91-line Kael Davan pack with phase-by-phase structure, prerequisite progression, and sample YAML
8. FriendArc state machine — 4-phase transitions with explicit trigger code and content dependencies
9. Wiki fix execution plan — 8 prioritized changes with per-file specifications for top-priority items
10. 37 new tickets + 10 updates — Full Sprint 3-5 roadmap across copy, server, client, and CI teams (NEW-12 killed in closing round)

What Remains Unresolved

• 3 NPC surnames awaiting Miri validation (Q-014)
• Interaction struct naming — trivial, resolved at implementation time (Q-015)
• 695 authored items scope validation (Q-017, informational)

Critical Path Unchanged

#261 (Dual Lens Authoring Guide) remains the single biggest blocker. Blocks 9 downstream tickets. Recommended: assign Paula, Sprint 3 P0, time-box 5 days.

---

Qatux, Round 2 tracking (updated with closing round resolutions). For the record: all 7 Round 2 responses cataloged. 4 lead decisions implemented + 3 closing round resolutions (Q-012: 16 keys, Q-013: server-authoritative pause, Q-016: verbs[] rename confirmed). 10 Round 1 disagreements resolved (6 changed to Paula, 3 kept Gestalt, 1 deferred by design). 20 decisions ready to formalize (D-042 through D-061: 13 new from Round 2 + closing, 7 carried from Round 1). 37 new tickets proposed + 10 updates (NEW-12 killed). 3 minor items remain open. Workshop complete.