Wiki Index

Last updated: 2026-07-08 | Total pages: 109 | Sources ingested: 96

Navigation: overview | Start Here | CONVENTIONS | README | dashboard

Entities

Anthropic API and Claude Platform (developing): The Anthropic API is the raw model surface beneath every Claude harness, and this brain currently verifies its position in the …
Anthropic (developing): Anthropic is the AI company behind the Claude models and product surfaces; in June 2026 it shipped Claude Fable 5 and Claude My…
Claude 5 Model Family (developing): The current Claude lineup as of 2026-07-07 is five pinned-snapshot models: Claude Fable 5 and Claude Mythos 5 at (path) per MTo…
Claude Agent SDK (developing): The Claude Agent SDK, renamed from the Claude Code SDK in September 2025, exposes the same tools, agent loop, and context manag…
Claude Beta Surface Agents (developing): This entity tracks the beta agents named in Anthropic product information for specific work surfaces, a Chrome browser agent an…
Claude Code (developing): Claude Code is Anthropic’s agentic coding harness, shipped as a CLI with VS Code, JetBrains, Desktop, and Web surfaces, and its…
Claude Cowork (developing): Claude Cowork is catalogued in this brain as Anthropic’s knowledge-work agent product, but none of the current claim packs veri…
Claude Fable 5 (developing): Claude Fable 5 is Anthropic’s most capable generally available model, a Mythos-class system launched 2026-06-09 with a 1M token…
Claude Mythos 5 (developing): Claude Mythos 5 is the trusted-access sibling of Claude Fable 5: the same underlying model offered to approved Project Glasswing partners…
Model Context Protocol (developing): MCP is an open JSON-RPC 2.0 protocol, current specification revision 2025-11-25, that standardizes how agents reach external to…
Project Glasswing (mature): Project Glasswing is Anthropic’s trusted-access defensive cybersecurity program for Mythos-class capability: it gives vetted defenders access to models and tooling…
System Prompt Export 2026-07 (developing): The System Prompt Export 2026-07 is this vault’s primary corpus: a 3826-line capture of the claude.ai consumer system prompt ru…
docs.claude.com (developing): docs.claude.com is Anthropic’s official Claude documentation site and this brain’s primary corroboration source; the three page…

Concepts

Adaptive Thinking and Thinking Mode (developing): Adaptive thinking is the always-on reasoning system of Fable 5, tuned only through the effort parameter, and the phrase “thinki…
Agent Skills System (developing): A skill is a folder with a SKILL.md entrypoint that loads through progressive disclosure, costing about 100 tokens until a matc…
Anthropic API in Artifacts (developing): claude.ai artifacts can call the Anthropic (path) endpoint directly with no API key, pinned to Sonnet 4 with max_tokens 1000, o…
Anthropic Runtime Reminders (developing): The claude.ai harness can inject six named reminder types mid-conversation when a classifier fires or another condition is met,…
Artifact Persistent Storage API (developing): claude.ai artifacts persist data across sessions through a window.storage key-value API with get, set, delete, and list methods…
Artifacts Usage Criteria (developing): claude.ai promotes content to an artifact when it is standalone and reusable, meaning any code over 20 lines or a text-heavy do…
Child Safety Rules (developing): The claude.ai system prompt embeds a critical_child_safety_instructions block inside refusal handling that sets a heightened-ca…
Citation Discipline (developing): Two citation regimes govern this brain: on claude.ai, Fable 5 must wrap every search-derived claim in cite tags and paraphrase …
Claude Character and Constitution (mature): Fable 5’s mentality is a trained value system, not a prompt overlay: the constitution fixes the priorit…
Claude Code Hooks (developing): Hooks are the deterministic layer of Claude Code: shell commands that fire at fixed lifecycle events and, unlike advisory CLAUD…
Claude Code Project Memory (developing): CLAUDE.md is Claude Code’s always-loaded instruction file, and it earns its context cost only when kept under 200 lines of fact…
Claude Code Settings and Permissions (developing): Claude Code settings resolve through a five-level hierarchy where managed org policy always wins, and permission rules merge ac…
Claude Code Subagents (developing): Subagents are Markdown-defined workers that each run in their own context window, so research, review, and verification stop co…
Claude Memory System (developing): The claude.ai memory system derives memories from past conversations in the background and binds Claude to strict framing, sens…
Computer Use File Handling Rules (developing): claude.ai’s computer-use container enforces a three-directory scheme, uploads in (path) scratch work in (path) which users neve…
Connector and MCP App Suggestions (developing): claude.ai routes third-party app needs through search_mcp_registry then suggest_connectors, and consumer-partner tools tagged […
Context Window Management (developing): Claude Fable 5 runs a 1M-token context window by default, and Anthropic’s primary management strategy is server-side compaction…
Copyright Compliance Rules (developing): The claude.ai harness enforces three hard copyright limits on Fable 5, quotes under 15 words, one quote per source, and zero re…
Core Search Behaviors (developing): The claude.ai harness gives Fable 5 a graduated search discipline: never search for timeless facts, one call for single facts, …
Evenhandedness Rules (developing): On claude.ai, Fable 5 treats persuasive requests as requests for the defenders’ best case rather than its own view, appends opp…
Extended Thinking Budgets (mature): Token-denominated thinking budgets are a pre-Fable-5 control: on Fable 5 adaptive thinking is always on, budget_tokens is rejec…
Fable 5 Dual-Use Safety Measures (developing): Fable 5 is the classifier-guarded release of the Mythos-class model: four refusal categories (cyber, bio, frontier_llm, reasoni…
Harness Refusal Handling (developing): The claude.ai system prompt opens its refusal_handling block with a permissive default, Claude “can discuss virtually any topic…
Headless Claude Code and CI (developing): Non-interactive mode, claude -p, is the official path for CI, scripts, and pipelines, and --bare is the recommended flag fo…
Image Search Decision Rules (developing): The claude.ai image search tool is governed by one enhance-understanding test, an explicit skip-list for text-shaped work, a le…
Knowledge Cutoff and Search Triggers (developing): Fable 5 on claude.ai treats end of January 2026 as its reliable knowledge boundary and is instructed to search without asking w…
Memory Injection Resistance (developing): The claude.ai memory system resists injected instructions structurally, not with a documented scanner: memory is written in the…
Model Selection for Agent Workloads (developing): For agent work in July 2026, default to Sonnet 5 at (path) escalate to Fable 5 at (path) only when the task needs frontier capa…
Mythos-Class Model Tier (developing): The Mythos class is Anthropic’s capability tier above Opus, launched 2026-06-09 as two variants of one underlying model: Claude…
Past Chats Tools (developing): conversation_search and recent_chats are claude.ai’s inline retrieval tools over past conversations, verified in the export’s 2…
Plan Mode and Permission Modes (developing): Claude Code ships six permission modes, and the high-leverage move is plan mode for research before changes, with bypassPermiss…
Prompt Caching Economics (developing): Cache reads bill at 0.1x base input price, a 90 percent discount on Fable 5 input tokens, so the economics reward stable prompt…
Responding to Mistakes and Criticism (developing): On claude.ai, Fable 5 owns its mistakes without over-apologizing, points unhappy users to the thumbs-down button, and may end t…
Sandbox Network and Filesystem Model (developing): Claude’s execution sandboxes differ sharply by surface: the claude.ai container gives Fable 5 a three-directory filesystem with…
System Prompt Design Logic (mature): The Fable 5 claude.ai system prompt is an engineered control system built on seven recurring design moves: checklists before ca…
Tone and Formatting Rules (mature): The claude.ai system prompt gives Fable 5 a warm, prose-first voice with minimum formatting by default, then layers channel pre…
Tool Search and Deferred Tools (developing): Deferred tool loading exists for context economy: keep tool definitions out of the context window until a task actually needs t…
User Preferences System (developing): claude.ai injects saved user preferences through a userPreferences tag that Fable 5 applies conservatively: behavioral preferen…
User Wellbeing Rules (developing): The claude.ai harness gives Fable 5 a detailed emotional-intelligence policy: no unrequested diagnostic labels, no method detai…
Visualizer Decision Ladder (developing): Before producing any visual, claude.ai walks a fixed four-step checklist, prose first, then a category-matching connected MCP t…
Writing Effective Tools for Agents (developing): Anthropic’s tool-design doctrine is to ship a few consolidated, workflow-shaped tools with evaluated descriptions and token-eff…

Flows

Brain Refresh Flow (developing): When a source’s refresh_due date in (path) passes, this brain stops trusting it: scan the ledger, re-fetch every stale URL, re-…
Claim Verification Flow (developing): Every claim in this brain survives an adversarial check before it is asserted: an independent second source confirms it or the …
Context Compaction Routine (developing): Treat the context window as a finite attention budget: compact deliberately at task boundaries with notes, persistent artifacts…
Corpus Ingestion Flow (developing): A new system prompt export enters this vault only through .raw: save it under a dated name, record its sha256 in (path) parse a…
Explore Plan Code Commit (developing): Explore Plan Code Commit is the officially recommended Claude Code loop: research in read-only plan mode, approve a written pla…
Multi-Agent Fan-Out Research Flow (developing): Fan out parallel worker agents from a lead orchestrator only for breadth-first research worth roughly 15x chat token spend; Ant…
Prompt Engineering Workflow for Fable 5 (developing): Prompt work on Fable 5 is empirical: define success criteria and tests before tuning, write clear instructions with their motiv…

Sources

Anthropic Engineering Blog Shelf (developing): This shelf catalogs the Anthropic engineering posts this brain cites, dated and annotated; the canonical agent-building sequenc…
Export Chapter Artifacts API and Citations (developing): Lines 3407-3826 of the export let artifacts call the Anthropic API with harness-handled auth pinned to a Sonnet 4 snapshot, def…
Export Chapter Computer Use and Search (developing): Lines 1027-1692 of the export mandate SKILL.md-first computer use, a three-directory file scheme, 20-line artifact thresholds, …
Export Chapter Memory and Preferences (developing): Lines 193-1025 of the export specify the consumer memory system, artifact persistent storage, MCP app suggestion etiquette, pas…
Export Chapter Product and Behavior (developing): Lines 1-191 of the claude.ai system prompt export define the 190000-token runtime budget, the Fable 5 and Mythos 5 product stor…
Export Chapter Tool Schemas (developing): Lines 1693-3405 of the export define 24 inline tools in JSONSchema, from computer use through consumer widgets to MCP plumbing,…
Fable Mythos 5 System Card (mature): The Claude Fable 5 and Claude Mythos 5 system card is the primary public evidence source for the shared-weight model’s capabilities…
Public System Prompt Copies (developing): Every known public copy of the Fable 5 system prompt, official and community, corroborates this vault’s 3,826-line capture of 2…
research-pack-2026-07-07 (mature): Every entry below was fetched and verified on the retrieved date; claims derived from them were adversarially checked (see (pat…

Platforms

Claude Code Platform (developing): Claude Code is Anthropic’s developer harness, reachable as a CLI, a VS Code extension, a JetBrains plugin, Desktop, and Web, do…
Claude Console and API Platform (developing): Claude Fable 5 is “generally available on the Claude API, Claude Platform on AWS, Amazon Bedrock” plus Google Cloud and Microso…
claude.ai Platform (developing): claude.ai is Anthropic’s consumer harness, and since June 9, 2026 it serves Claude Fable 5 to Pro, Max, Team, and Enterprise su…

Accounts

Anthropic Plans and Access Tiers (developing): Claude Fable 5 ships to the paid subscription tiers, Pro, Max, Team, and Enterprise, on both claude.ai and Claude Code, while C…

Decisions

Confidence Tag Policy (developing): Every domain claim in this vault carries exactly one confidence tag assigned under five fixed rules from (path) which privilege…
Corpus Scope Decision (developing): The System Prompt Export 2026-07 is a June 9, 2026 capture of the claude.ai consumer system prompt, not Claude Code, so every c…
Frontmatter Dialect Decision (developing): Every note in this vault uses one merged frontmatter dialect, the claude-obsidian status vocabulary plus house enrichment field…

Deliverables

Claude Code Quickstart for Fable 5 (developing): Install Claude Code from the official per-surface guides, seed a short CLAUDE.md with (path) start work in plan mode, tighten p…
Fable 5 Best Practices Cheat Sheet (developing): The fifteen highest-leverage verified practices for operating Claude Fable 5 across model choice, prompting, context, and safet…
Fable Mythos Operating Doctrine (mature): Operate Fable 5 as the general-access Mythos-class workhorse and Mythos 5 as a restricted, monitored, trusted-access capability…
Prompting Fable 5 One-Pager (developing): Prompt Fable 5 with clear goals, stated motivation, and XML structure, let adaptive thinking manage depth through the effort pa…

Questions

Fable 5 Pricing and Rate Limits (developing): Pricing is fully documented at (path) per MTok with a 90% caching discount, but numeric rate limits for Fable 5 remain unverifi…
Mythos 5 Access Criteria (developing): The gate is documented (invitation-only Project Glasswing membership via an account team) but the actual approval criteria behi…
Which Export Rules Bind Claude Code (developing): Unresolved: the export is a claude.ai capture, so none of its rules are proven to bind Claude Code, whose behavior is documente…

Gaps

Export Omits Claude Code Harness (developing): The System Prompt Export 2026-07 contains no Claude Code system prompt, tool schemas, or harness rules, so every Claude Code cl…
No Public Fable 5 Benchmark Data (mature): Resolved 2026-07-07: the Fable/Mythos 5 system card now provides public evaluation evidence, though broad general-purpose benchmark…

Experiments

Tool Search Behavior Probe (seed): A designed, not yet run, probe testing whether deferred tools in the claude.ai harness stay name-only until a tool_search call …

Fable Brain

Explorer

Fable Brain

Wiki Index

Entities

Concepts

Flows

Sources

Platforms

Accounts

Decisions

Deliverables

Questions

Gaps

Experiments

Meta

Graph View

Table of Contents

Backlinks