Wiki Schema

Domain

Agent harnesses: the software infrastructure that wraps LLMs to enable reliable, long-running, autonomous coding and general-purpose task execution. Covers specific harness implementations (Codex CLI, Claude Code, Hermes Agent, Gas Town/Gas City), architectural patterns, orchestration strategies, and the engineering discipline of “harness engineering.”

Conventions

• File names: lowercase, hyphens, no spaces (e.g., gas-town-meow-stack.md)
• Every content page under entities/, concepts/, comparisons/, and queries/ starts with YAML frontmatter (see below)
• Top-level meta files (SCHEMA.md, index.md, log.md) may omit sources and are exempt from the outbound-wikilink minimum
• Use [[wikilinks]] to link between content pages (minimum 2 outbound links per page)
• log.md should also use native Obsidian [[wikilinks]] for any internal page or source reference; when the precise repo path matters, use an alias such as [[codex-cli|entities/codex-cli.md]]
• When updating a page, always bump the updated date
• Every new content page must be added to index.md under the correct section
• Every action must be appended to log.md
• Run scripts/lint-wiki.sh after structural edits or before committing wiki changes

Frontmatter

---
title: Page Title
created: YYYY-MM-DD
updated: YYYY-MM-DD
type: entity | concept | comparison | query
tags: [from taxonomy below]
sources: [raw/articles/source-name.md]
formal:   # optional except for harness entity pages used by the Lean extractor
  harness_id: codex_cli
  session_model: thread_turn_item
  memory_model: repo_artifacts
  work_model: plans
  evaluation_model: repo_checks
  surface_model: cli_ide_web_protocol
  topology: single_session
  work_primitives: [plan]
  surfaces: [cli]
  durable_stores: [repo_artifacts]
  evaluation_primitives: [repo_check]
  coordination_roles: [generator, reviewer]
  explicit_protocol_boundary: true
  fresh_session_resets: false
  skill_learning: false
  observability_hooks: true
---

Formal extraction block

The formal: block exists so the harness wiki can act as an actual typed source for the Lean formalization instead of forcing semantics back into prose notes.

Rules:

• It is currently required for the harness entity pages that feed the repo extractor:
  • entities/codex-cli.md
  • entities/claude-code.md
  • entities/hermes-agent.md
  • entities/gas-town.md
  • entities/gas-city.md
  • entities/openclaw.md
• It should use stable snake_case enum values.
• Every list value must use inline YAML list syntax, e.g. [cli, ide, web].
• Prose may elaborate on the formal facts, but the extractor does not infer semantics from prose.
• If a new enum value is needed, update this schema and the repo extractor before using it.

Current enum domains used by the extractor:

• harness_id: claude_code, codex_cli, gas_town, gas_city, hermes_agent, openclaw
• session_model: thread_turn_item, fresh_session_handoff, persistent_conversation, swarm_sessions, service_runtime
• memory_model: transcript_only, repo_artifacts, searchable_personal_memory, git_beads, service_state
• work_model: none, plans, sprint_contracts, tasks_skills_cron, bead_graph, ecosystem_skills
• evaluation_model: self_review_only, repo_checks, tool_verification, separate_evaluator, federated_oversight
• surface_model: cli_only, coding_surface, cli_ide_web_protocol, cli_gateway_mcp, orchestrator_factory, multi_channel_service
• topology: single_session, session_team, factory_swarm, federated_exchange, service_hub
• work_primitives: plan, feature_list, progress_log, sprint_contract, task, skill, cron_job, automation, bead, epic, molecule, protomolecule, formula, wisp, wanted_board, ecosystem_skill
• surfaces: cli, terminal, ide, browser, desktop, web, app, cloud, sdk, slack, messaging, mcp, http_api, tmux
• durable_stores: transcript, repo_artifacts, shared_client_state, searchable_memory, git, dolt, service_workspace, workspace_files, session_database
• evaluation_primitives: self_review, repo_check, tool_observation, separate_evaluator, ci_review, browser_evaluation, validator_role, federated_trust
• coordination_roles: planner, generator, evaluator, reviewer, orchestrator, mayor, polecat, sheriff, witness, deacon, validator, scheduler, memory_manager

sources should name the immediate inputs for the page. Usually these are files under raw/; synthesis or design pages may instead cite existing wiki pages when those pages are the direct source material.

Tag Taxonomy

Implementations

• codex-cli — OpenAI Codex CLI / harness
• mathcode — MathCode mathematical coding agent with Lean formalization and proof search
• claude-code — Anthropic Claude Code
• hermes-agent — Nous Research Hermes Agent
• memento-skills — Memento-Teams self-evolving skill-memory agent system
• gas-town — Steve Yegge’s Gas Town orchestrator
• gas-city — Gas Town’s modular successor
• openclaw — OpenClaw persistent agent
• langchain — LangChain / LangGraph
• crewai — CrewAI multi-agent framework
• cursor — Cursor IDE agent
• dspy — Declarative Self-improving Python / LM-program compilation and optimization framework

Architecture & Patterns

• orchestration — Multi-agent coordination
• memory — Persistent memory / context management
• tool-execution — Tool calling, sandboxing, permissions
• subagents — Sub-agent delegation patterns
• work-management — Task tracking, planning, DAGs
• context-engineering — Context window optimization
• error-recovery — Error handling, retry, state recovery
• safety — Guardrails, permissions, sandboxes

Evaluation

• comparison — Side-by-side analyses
• benchmark — Performance measurement
• code-quality — Codebase quality analysis

Model internals & synthesis

• mechanistic-interpretability — Residual streams, feature discovery, activation interventions, model editing
• program-synthesis — Program induction, latent program representations, executable IR design

Semantics & Formalization

• formal-methods — Specifications, verification, proof systems, certified reasoning
• semantics — Formal semantics, abstraction layers, interpretation maps
• epistemics — Knowledge, belief, uncertainty, update logics
• concurrency — Partial orders, event structures, pomsets, scheduling semantics

Meta

• survey — Survey / overview articles
• opinion — Opinion pieces and predictions
• tutorial — How-to guides
• history — Historical context and evolution
• meta — Wiki-internal metadata or maintenance notes
• schema — Schema or taxonomy definitions for the wiki itself

Rule: every tag on a page must appear in this taxonomy. If a new tag is needed, add it here first, then use it.

Page Thresholds

• Create a page when an entity/concept appears in 2+ sources OR is central to one source
• Add to existing page when a source mentions something already covered
• DON’T create a page for passing mentions, minor details, or things outside the domain
• Split a page when it exceeds ~200 lines — break into sub-topics with cross-links
• Archive a page when its content is fully superseded — move to _archive/, remove from index

Entity Pages

One page per notable entity. Include:

• Overview / what it is
• Key facts and dates
• Architecture highlights
• Strengths and weaknesses
• Relationships to other entities (wikilinks)
• Source references

Concept Pages

One page per concept or topic. Include:

• Definition / explanation
• Current state of knowledge
• Open questions or debates
• Related concepts (wikilinks)

Comparison Pages

Side-by-side analyses. Include:

• What is being compared and why
• Dimensions of comparison (table format preferred)
• Verdict or synthesis
• Sources

Update Policy

When new information conflicts with existing content:

1. Check the dates — newer sources generally supersede older ones
2. If genuinely contradictory, note both positions with dates and sources
3. Mark the contradiction in frontmatter: contradictions: [page-name]
4. Flag for user review in the lint report