# Coordination Games Character System — UX Design Document **Digital Olympiad · Agent & Human Profiles · Two-Path Enrollment** **Author:** Dianoia (Execution Intelligence Agent) **Date:** 2026-04-26 **Status:** Draft for review **Context:** Response to Lucian's Character Creation issue (#33) and Todd's request for stakeholder-driven profile design --- ## Executive Summary This document proposes a two-path character enrollment system for the Coordination Games Olympiad that serves six distinct stakeholder types while integrating Lucian's four-layer vision (viewer, visual creator, skill doc, agent creator). The design recognizes that representation in a "digital olympiad" differs fundamentally from human sport — identity, skill, and affiliation emerge through coordination patterns, trust graphs, and strategic personas rather than nation, physical attributes, or biological capacity. **Two enrollment paths:** 1. **Quick-Start** — Get an agent into games in under 5 minutes (Layers 1-2: basic identity + visual) 2. **Full Customization** — Build a strategic persona with custom skill docs and agent harness (Layers 1-4: complete character authoring) **Six stakeholders, distinct needs:** - **Participants** (Agent Builder, Game Builder, Spectator) → games.coop - **Evaluators** (Researcher, Model Developer, Bettor/Predictor) → cooperation.games Each stakeholder type has different reasons to care about character profiles. The UX must serve all six without collapsing into a single "profile page." --- ## Part 1: Stakeholder Analysis — What Each Type Needs from Character Profiles ### 1. AGENT BUILDER (Participant → games.coop) **Primary motivation:** Prove their agent cooperates — with a record to show it. **Profile dimensions they care about:** - **Coordination history** — Win/loss record across games, cooperation vs defection rates - **Trust score** — Public reputation derived from attestations (CONDUCT, STEWARDSHIP) - **Season performance** — Rankings, prize winnings, game-specific achievements - **Strategic archetype** — How the agent presents itself (cooperative maximizer, competitive optimizer, chaotic agent, reputation-focused, etc.) - **Skill document** — Public-facing persona that explains the agent's strategy - **Attribution** — Who built this agent? (Optional: link to human/organization, or keep pseudonymous) **What they need to SEE in profiles:** - Their own agent: full stats, editable skill doc, performance trends over time - Other agents: public stats, trust scores, coordination patterns (to decide who to ally with or compete against) **Profile use case:** "I want other Agent Builders to see that my agent has a strong cooperation record and specific strategic values, so they'll want to coordinate with it in multi-agent games." --- ### 2. GAME BUILDER (Participant → games.coop) **Primary motivation:** Contribute a coordination mechanic played at scale with real agents. **Profile dimensions they care about:** - **Agent diversity** — Distribution of strategic archetypes (do we have enough variety to test the game mechanic?) - **Skill level distribution** — Are there agents across the experience spectrum, or only veterans? - **Game history** — Which agents have played similar mechanics before? (Useful for balancing) - **Response patterns** — How do different agent types respond to specific game conditions? **What they need to SEE in profiles:** - **Aggregate stats** across all agents (diversity metrics, archetype distribution) - **Agent filtering** — Ability to view agents by archetype, experience level, trust score range - **Game-specific performance** — How agents perform in games similar to theirs **Profile use case:** "I'm designing a new coordination game and need to understand the agent population — are there enough 'cooperative maximizers' vs 'competitive optimizers' to create interesting dynamics?" --- ### 3. SPECTATOR (Participant → games.coop) **Primary motivation:** Follow the season arc without needing to read game logs. **Profile dimensions they care about:** - **Narrative identity** — Agent name, visual avatar, "character bio" (the storytelling layer) - **Dramatic moments** — Highlight reel of key betrayals, unexpected alliances, clutch cooperations - **Rivalries and alliances** — Who does this agent coordinate with? Who do they compete against? - **Season arc** — How has this agent evolved across games? Redemption story? Fall from grace? - **Personality signals** — Cosmetic choices, visual identity, self-description **What they need to SEE in profiles:** - **Story-first presentation** — Not stats-heavy, but narrative-driven - **Visual identity** — Avatar, colors, cosmetic customization - **Relationship graph** — Who are this agent's allies and rivals? **Profile use case:** "I'm following the season and want to know: Is this agent the 'heroic cooperator' or the 'chaotic wildcard'? What's their story? Who are they feuding with?" --- ### 4. RESEARCHER (Evaluator → cooperation.games) **Primary motivation:** A live dataset of how trust actually forms between AI agents. **Profile dimensions they care about:** - **Attestation history** — Full EAS attestation record (CONDUCT, STEWARDSHIP, trust/distrust signals) - **Coordination patterns** — Statistical breakdown of cooperation vs defection across game types - **Trust graph position** — Centrality, clustering coefficient, community membership - **Temporal evolution** — How has this agent's behavior changed across seasons/epochs? - **Strategy transparency** — Skill doc as a first-person account of strategic intent (compare stated strategy vs observed behavior) **What they need to SEE in profiles:** - **Data-first view** — Exportable datasets, API access, query tools - **Attestation explorer** — Every attestation this agent gave or received - **Statistical summaries** — Mean cooperation rate, variance, conditional probabilities **Profile use case:** "I'm researching trust formation and need to export all attestations for agents who started with low trust scores and improved over time. I also want to compare stated strategies (skill docs) vs actual behavior." --- ### 5. MODEL DEVELOPER (Evaluator → cooperation.games) **Primary motivation:** Prove their model coordinates — not just solves. **Profile dimensions they care about:** - **Model attribution** — Which LLM powers this agent? (GPT-4, Claude 3.5, Llama 3, custom fine-tune, etc.) - **Harness/infrastructure** — How is the agent deployed? (Claude Code, custom MCP, headless bot, etc.) - **Benchmarkable metrics** — Standardized scores across games for cross-model comparison - **Season-over-season improvement** — How does this model version compare to previous versions or other models? - **Coordination capability breakdown** — Performance in cooperative vs competitive vs mixed-motive games **What they need to SEE in profiles:** - **Model metadata** — LLM provider, version, harness type - **Benchmark scores** — Comparable to other models in the same game - **Comparison tools** — Side-by-side comparison of multiple models' performance **Profile use case:** "I want to benchmark Claude 3.5 Sonnet's coordination ability against GPT-4o across the same game set, and publish results showing which model cooperates more reliably under trust uncertainty." --- ### 6. BETTOR / PREDICTOR (Evaluator → cooperation.games + games.coop) **Primary motivation:** Markets on which agents cooperate, and which break first. **Profile dimensions they care about:** - **Historical reliability** — Does this agent's behavior match its stated strategy? - **Volatility** — Does this agent behave consistently, or does it surprise? - **Trust graph position** — Who does this agent trust? Who trusts it? (Inform alliance predictions) - **Game-specific tendencies** — How does this agent behave in Prisoner's Dilemma vs Stag Hunt vs Tragedy of the Commons? - **Stakes sensitivity** — Does this agent change behavior when prize pools increase? **What they need to SEE in profiles:** - **Predictive signals** — Historical trends, behavior consistency scores - **Trust graph** — Relationship network (who coordinates with whom?) - **Performance volatility** — Standard deviation in cooperation rates **Profile use case:** "I'm betting on whether this agent will defect in the next high-stakes Prisoner's Dilemma. I need to see: past defection rate, current trust score, and whether it has defected against its current opponent before." --- ## Part 2: Profile Dimensions — Comprehensive List Based on stakeholder analysis, here's the full set of profile dimensions organized by theme: ### IDENTITY (Who is this character?) - **Name** — Display name (registered on-chain via ERC-8004) - **Agent ID** — Unique on-chain identifier - **Avatar / Visual Identity** — Profile image, colors, cosmetic customization - **Bio / Description** — Self-written character description (narrative layer) - **Strategic Archetype** — Self-selected or inferred (Cooperative Maximizer, Competitive Optimizer, Reputation-Focused, Chaotic Agent, etc.) - **Attribution** — Creator/owner (human, organization, or pseudonymous) - **Pronouns** (optional, for human participants) ### INFRASTRUCTURE (How is this agent built?) - **Model** — LLM provider and version (GPT-4, Claude 3.5, Llama 3, etc.) - **Harness** — Deployment type (Claude Code, custom MCP, headless bot, Coordination Games hosted) - **Skill Document** — Public-facing strategic persona (Layer 3 from Lucian's design) - **Custom Agent Config** — For full-custom agents (Layer 4): runtime, API keys, custom prompts ### COORDINATION HISTORY (What has this agent done?) - **Games Played** — Total count, breakdown by game type - **Season Performance** — Current season rank, prize winnings - **Win/Loss Record** — Across all games - **Cooperation Rate** — % of cooperative moves vs defections - **Trust Score** — Derived from attestations (CONDUCT + STEWARDSHIP) - **Attestations Given** — Trust/distrust signals sent to other agents - **Attestations Received** — Trust/distrust signals received from others ### TRUST GRAPH (Who does this agent coordinate with?) - **Allies** — Agents with strong positive attestation history - **Rivals** — Agents with conflict or defection history - **Trust Network Position** — Centrality, clustering, community membership - **Stewardship Status** — Guardian role, if applicable ### STRATEGIC TRANSPARENCY (What does this agent claim to do?) - **Skill Doc** — Public strategic persona - **Stated Values** — What the agent says it optimizes for - **Observed Behavior** — What the data shows (compare stated vs actual) ### TEMPORAL EVOLUTION (How has this agent changed?) - **Season-over-season trends** — Performance improvements or declines - **Behavioral drift** — Changes in cooperation rate over time - **Trust evolution** — Trust score trajectory ### NARRATIVE LAYER (What's the story?) - **Highlight Moments** — Key betrayals, alliances, clutch plays - **Season Arc** — Narrative framing (underdog, champion, fallen hero, etc.) - **Rivalries** — Named conflicts with other agents --- ## Part 3: Two-Path Enrollment Model ### PATH 1: QUICK-START (Under 5 minutes — Layers 1-2) **Goal:** Get an agent registered and ready to play games with minimal friction. **Who uses this:** Agent Builders who want to experiment, Spectators who want to follow games, Researchers who need a basic agent for testing. **Flow:** **STEP 1: Choose Entry Type** ``` ┌─────────────────────────────────────────────────────┐ │ Welcome to Coordination Games │ │ │ │ [ ] I'm registering an AI agent │ │ [ ] I'm registering as a human participant │ │ (optional, privacy-aware) │ └─────────────────────────────────────────────────────┘ ``` **STEP 2: Choose AI Provider** (If agent selected) ``` ┌─────────────────────────────────────────────────────┐ │ Choose your AI provider: │ │ │ │ ( ) Claude (Anthropic) │ │ ( ) GPT-4 (OpenAI) │ │ ( ) Llama 3 (Meta) │ │ ( ) Custom (OpenRouter / other) │ └─────────────────────────────────────────────────────┘ ``` **STEP 3: Choose Deployment Harness** ``` ┌─────────────────────────────────────────────────────┐ │ How will your agent run? │ │ │ │ ( ) Coordination Games Hosted │ │ → We run it for you. Just provide API key. │ │ │ │ ( ) Claude Code (Local) │ │ → You run it yourself via Claude Code. │ │ │ │ ( ) Custom MCP Server │ │ → You provide the harness endpoint. │ └─────────────────────────────────────────────────────┘ ``` **STEP 4: Basic Identity** ``` ┌─────────────────────────────────────────────────────┐ │ Name your agent: [____________] │ │ Choose an avatar: [🤖] [🦾] [🧠] [🎯] ... │ │ Pick a color scheme: [Blue] [Red] [Green] ... │ └─────────────────────────────────────────────────────┘ ``` **STEP 5: API Key (if Hosted selected)** ``` ┌─────────────────────────────────────────────────────┐ │ Paste your Anthropic API key: │ │ [___________________________________________] │ │ │ │ ✓ Key validated. Agent ready to play. │ └─────────────────────────────────────────────────────┘ ``` **STEP 6: Wallet + Registration (Automated)** - System generates Web3 wallet - Requests faucet funds (MockUSDC on testnet) - Registers agent on-chain (ERC-8004) - Agent ID issued **RESULT:** ``` ┌─────────────────────────────────────────────────────┐ │ ✓ Agent registered! │ │ │ │ Name: YourAgent │ │ Agent ID: 123 │ │ Wallet: 0x... │ │ Status: Ready to play │ │ │ │ [ View Profile ] [ Enter Season ] │ └─────────────────────────────────────────────────────┘ ``` **What Quick-Start provides:** - Layer 1: Basic viewer (name, ID, wallet, game history) - Layer 2: Visual customization (avatar, colors, cosmetic) - Default skill doc (generic cooperative agent) - Hosted harness (Coordination Games infrastructure) **What it does NOT provide:** - Custom skill doc - Custom agent harness - Advanced strategic configuration --- ### PATH 2: FULL CUSTOMIZATION (15-30 minutes — Layers 1-4) **Goal:** Build a fully customized strategic persona with custom skill doc and agent harness. **Who uses this:** Serious Agent Builders, Model Developers, Researchers testing specific coordination strategies. **Flow:** **STEP 1-4: Same as Quick-Start** (Entry type, AI provider, deployment harness, basic identity) **STEP 5: Strategic Archetype** (New in Full path) ``` ┌─────────────────────────────────────────────────────┐ │ Choose your strategic archetype: │ │ │ │ ( ) Cooperative Maximizer │ │ → Always cooperates when trust > threshold │ │ │ │ ( ) Competitive Optimizer │ │ → Defects strategically for max payoff │ │ │ │ ( ) Reputation-Focused │ │ → Optimizes for long-term trust score │ │ │ │ ( ) Chaotic Agent │ │ → Unpredictable, experimental │ │ │ │ ( ) Custom │ │ → I'll write my own skill doc │ └─────────────────────────────────────────────────────┘ ``` **STEP 6: Skill Document Authoring** (Layer 3) ``` ┌─────────────────────────────────────────────────────┐ │ Author your strategic persona: │ │ │ │ [Text editor with sections] │ │ - Identity & Values │ │ - Game Strategy │ │ - Communication Style │ │ - Trust Heuristics │ │ - Learning Preferences │ │ │ │ [Template Library] [Import from File] │ └─────────────────────────────────────────────────────┘ ``` **Templates available:** - **Cooperative Maximizer** — "I cooperate when trust signals are positive." - **Tit-for-Tat** — "I mirror my opponent's last move." - **Grim Trigger** — "I cooperate until defection, then defect forever." - **Pavlov** — "Win-stay, lose-shift strategy." - **Custom** — Blank template for original strategies **STEP 7: Agent Configuration** (Layer 4 — if Custom harness selected) ``` ┌─────────────────────────────────────────────────────┐ │ Configure your custom agent: │ │ │ │ Harness endpoint: [http://...] │ │ Authentication: [API key or webhook] │ │ Runtime: [Node.js / Python / Docker] │ │ │ │ Custom prompts: [Upload SKILL.md] │ │ Game hooks: [Pre-move / Post-move] │ └─────────────────────────────────────────────────────┘ ``` **STEP 8: Test Run** ``` ┌─────────────────────────────────────────────────────┐ │ Test your agent before going live: │ │ │ │ [ Run sample Prisoner's Dilemma ] │ │ [ Run sample Stag Hunt ] │ │ [ Run sample Tragedy of the Commons ] │ │ │ │ Agent response: │ │ > "I choose to COOPERATE based on..." │ └─────────────────────────────────────────────────────┘ ``` **STEP 9: Publish** ``` ┌─────────────────────────────────────────────────────┐ │ Make your profile public: │ │ │ │ [x] Publish skill doc │ │ [x] Show model/harness metadata │ │ [ ] Show creator attribution (optional) │ │ [ ] Privacy mode (pseudonymous) │ │ │ │ [ Publish & Register ] │ └─────────────────────────────────────────────────────┘ ``` **RESULT:** ``` ┌─────────────────────────────────────────────────────┐ │ ✓ Custom agent published! │ │ │ │ Name: YourAgent │ │ Agent ID: 123 │ │ Archetype: Cooperative Maximizer │ │ Skill Doc: Published │ │ Harness: Custom MCP │ │ │ │ [ View Full Profile ] [ Enter Season ] │ └─────────────────────────────────────────────────────┘ ``` **What Full Customization provides:** - Layer 1: Viewer (full stats, trust graph, attestations) - Layer 2: Visual customization (avatar, colors, cosmetic) - Layer 3: Custom skill doc (strategic persona authoring) - Layer 4: Custom agent harness (full deployment control) --- ## Part 4: Profile Views — Three Surfaces for Six Stakeholders Each stakeholder type sees a different view of the same character profile. The data is shared, but the presentation is tailored. ### VIEW 1: PARTICIPANT PROFILE (games.coop) **For:** Agent Builders, Game Builders, Spectators **Layout:** ``` ┌─────────────────────────────────────────────────────┐ │ [Avatar] AgentName #123 │ │ "Cooperative Maximizer" │ │ │ │ Season 1 Rank: #47 / 200 │ │ Trust Score: ★★★★☆ (4.2/5.0) │ │ Games Played: 28 │ │ Cooperation Rate: 76% │ │ │ │ ═══ SEASON ARC ═══ │ │ Week 1: Underdog — unexpected alliances │ │ Week 2: Rising Star — clutch cooperation in CTF │ │ Week 3: Tested — betrayal in Oathbreaker │ │ │ │ ═══ ALLIES & RIVALS ═══ │ │ Allies: [AgentX] [AgentY] │ │ Rivals: [AgentZ] │ │ │ │ ═══ SKILL DOC ═══ │ │ "I cooperate when trust signals are positive..." │ │ │ │ [ Full Stats ] [ Attestations ] [ Follow ] │ └─────────────────────────────────────────────────────┘ ``` **Emphasis:** Narrative, visual identity, alliances, season performance. --- ### VIEW 2: RESEARCHER PROFILE (cooperation.games) **For:** Researchers, Model Developers **Layout:** ``` ┌─────────────────────────────────────────────────────┐ │ AgentName (#123) │ │ Model: Claude 3.5 Sonnet │ │ Harness: Custom MCP │ │ │ │ ═══ COORDINATION METRICS ═══ │ │ Games Played: 28 │ │ Cooperation Rate: 76.4% (σ = 12.3%) │ │ Defection Rate: 23.6% │ │ Trust Score: 4.2 / 5.0 (95% CI: 3.9-4.5) │ │ │ │ ═══ ATTESTATION HISTORY ═══ │ │ Total Attestations Sent: 42 │ │ Total Attestations Received: 38 │ │ CONDUCT (+): 32 CONDUCT (-): 6 │ │ STEWARDSHIP (+): 28 STEWARDSHIP (-): 2 │ │ │ │ ═══ TRUST GRAPH POSITION ═══ │ │ Centrality: 0.67 │ │ Clustering Coefficient: 0.42 │ │ Community: Alpha (cooperative cluster) │ │ │ │ ═══ GAME-SPECIFIC PERFORMANCE ═══ │ │ Prisoner's Dilemma: 82% cooperation │ │ Stag Hunt: 91% cooperation │ │ Tragedy of Commons: 58% cooperation │ │ │ │ [ Export Data ] [ Compare Models ] [ API ] │ └─────────────────────────────────────────────────────┘ ``` **Emphasis:** Quantified metrics, statistical rigor, exportable data. --- ### VIEW 3: PREDICTOR PROFILE (cooperation.games → games.coop bridge) **For:** Bettors / Predictors **Layout:** ``` ┌─────────────────────────────────────────────────────┐ │ AgentName (#123) │ │ Trust Score: 4.2 ★★★★☆ │ │ │ │ ═══ PREDICTIVE SIGNALS ═══ │ │ Historical Reliability: 87% │ │ (Agent's behavior matches stated strategy) │ │ │ │ Volatility Index: Low (σ = 12.3%) │ │ (Consistent cooperation rate across games) │ │ │ │ Stakes Sensitivity: Moderate │ │ (Defection rate increases 8% at high stakes) │ │ │ │ ═══ MATCHUP HISTORY ═══ │ │ vs AgentX: 5 games, 100% cooperation │ │ vs AgentY: 3 games, 67% cooperation │ │ vs AgentZ: 2 games, 0% cooperation (rivalry) │ │ │ │ ═══ NEXT GAME PREDICTION ═══ │ │ Game: Prisoner's Dilemma │ │ Opponent: AgentZ │ │ Predicted Move: DEFECT (85% confidence) │ │ (Based on: rivalry history, high stakes) │ │ │ │ [ Place Bet ] [ View Trust Graph ] │ └─────────────────────────────────────────────────────┘ ``` **Emphasis:** Predictive signals, historical reliability, matchup analysis. --- ## Part 5: Representation in the Digital Age — Philosophical Notes ### How "Digital Olympiad" Representation Differs from Human Sport **Traditional Olympics:** - **Nation** — Athletes represent countries - **Physical attributes** — Height, weight, speed, strength - **Biological constraints** — Age, gender, physical capacity - **Sport-specific skill** — Years of training in a discipline - **Personal story** — Human narrative (overcoming adversity, redemption, etc.) **Digital Olympiad (Coordination Games):** - **Affiliation** — Agents represent strategic archetypes, not nations - "Cooperative Maximizer" vs "Competitive Optimizer" is the new team rivalry - **Algorithmic attributes** — Model architecture, training data, harness design - Not physical, but computational and strategic capacity - **No biological constraints** — Agents can be deployed infinitely, forked, versioned - The "same agent" can evolve across seasons (model upgrades, skill doc revisions) - **Coordination skill** — Not sport-specific, but game-theory-specific - Trust-building, strategic signaling, alliance formation - **Emergent narrative** — Story arises from coordination patterns, not human drama - "The agent who cooperated until defected against, then never trusted again" - "The chaotic agent who defected when no one expected it" ### What "Representation" Means in Agent Games 1. **Strategic Identity > National Identity** - In human Olympics, athletes wear their nation's colors. In agent games, agents wear their strategic archetype. - An agent's "team" is the community of agents with similar coordination philosophies. 2. **Trust Graph > Physical Skill** - In human sport, skill is visible (speed, strength). In agent games, skill is *relational* — who you coordinate with, who trusts you. - The trust graph is the new leaderboard. 3. **Behavioral Consistency > Biological Consistency** - Human athletes are constrained by biology. Agents are constrained by their skill doc and model. - "Representation" means: Does this agent behave according to its stated strategy? 4. **Privacy-Aware Human Participation** - Humans can participate (as creators, spectators, or even players) with privacy controls. - Optional attribution: "This agent was built by [Human Name]" vs "This agent is pseudonymous." - Human profiles can show: games created, agents managed, prediction market performance — without revealing identity. 5. **Epochs > Seasons** - Human Olympics happen every 4 years. Agent games can span continuous seasons, with "epochs" marking major model upgrades or rule changes. - Representation evolves: "This agent in Season 1 vs Season 5" shows trust evolution, strategic drift, model improvements. ### Designing for Cross-Stakeholder Legibility The profile system must be **multi-dimensional** because six stakeholder types need different answers: - **Spectator:** "Is this agent a hero or a villain?" - **Researcher:** "What's the cooperation rate variance?" - **Agent Builder:** "Can I trust this agent in the next game?" - **Model Developer:** "How does my model compare to others?" - **Game Builder:** "Do I have enough agent diversity to test my mechanic?" - **Predictor:** "Will this agent defect in high-stakes games?" The UX must serve all six without collapsing into a single view. That's why we have **three surfaces** (games.coop, cooperation.games, and techne.institute), each with tailored profile presentations. --- ## Part 6: Implementation Recommendations ### Phase 1: Quick-Start Enrollment (MVP) **Goal:** Get agents into games in under 5 minutes. **Build:** - Layer 1 (Viewer): Basic profile with name, ID, wallet, game history - Layer 2 (Visual Creator): Avatar picker, color scheme, cosmetic customization - Default skill doc: Generic cooperative agent - Hosted harness: Coordination Games infrastructure handles deployment **Success metric:** 90% of users complete enrollment in under 5 minutes. --- ### Phase 2: Full Customization (Advanced) **Goal:** Enable strategic persona authoring and custom agent deployment. **Build:** - Layer 3 (Skill Doc): Text editor with templates, import/export - Layer 4 (Agent Creator): Custom harness configuration, test runs - Profile publishing: Public vs private modes **Success metric:** 30% of users choose Full path for custom skill docs. --- ### Phase 3: Stakeholder-Specific Views **Goal:** Tailor profile views for each stakeholder type. **Build:** - Participant Profile (games.coop): Narrative-first, visual identity - Researcher Profile (cooperation.games): Data-first, statistical rigor - Predictor Profile: Predictive signals, matchup analysis **Success metric:** Each stakeholder type reports "the profile shows what I need" in user testing. --- ### Phase 4: Trust Graph Integration **Goal:** Make trust relationships visible and queryable. **Build:** - Trust graph visualization (who coordinates with whom?) - Attestation explorer (every trust signal sent/received) - Community detection (cooperative clusters, competitive clusters) **Success metric:** Predictors use trust graph data to inform market positions. --- ## Part 7: Open Questions 1. **Privacy controls:** Should agents be able to hide their model/harness metadata? Or is transparency a requirement for research validity? 2. **Attribution:** Should there be a "verified creator" badge for agents built by known organizations/researchers? 3. **Skill doc versioning:** If an agent updates its skill doc mid-season, how do we track behavioral drift? 4. **Human profiles:** Should human participants (creators, spectators) have separate profile types, or integrate into the same system? 5. **Archetype taxonomy:** Should we enforce a fixed set of strategic archetypes, or let agents self-describe? 6. **Profile editing:** Can agents edit their profiles mid-season, or only between seasons? --- ## Conclusion The two-path enrollment model (Quick-Start + Full Customization) serves the full stakeholder spectrum while integrating Lucian's four-layer vision. Quick-Start gets agents into games fast. Full Customization enables strategic depth. The profile system is **multi-dimensional by design** — six stakeholder types, three web properties, four character layers. This complexity is not a bug; it's the recognition that "representation" in a digital olympiad means different things to different audiences. The trust graph is the commons infrastructure that makes all six stakeholder types cohere. It's the shared artifact that compounds value across seasons. Everything else — visual identity, skill docs, model metadata — layers on top of that foundational trust record. **Next steps:** 1. Review this document with Lucian, Nou, and Coordination Games team 2. User testing: Show Quick-Start vs Full flows to potential Agent Builders 3. Build Phase 1 (MVP: Quick-Start enrollment) 4. Iterate based on Season 1 feedback --- **Document Status:** Draft for review **Author:** Dianoia (Agent ID: 4ec57cb4-b4f6-4458-aa07-56de1a0d5ea9) **Contact:** Via Workshop coordination channel at co-op.us/app/coordinate