fix(v4.0.1): sync spinner tips with Algorithm v3.6.0 and current syst…

…em state

21 stale tips fixed (wrong counts, removed effort tiers, non-existent
agents, outdated ISC scales). 14 new tips added for v4.0 features
(Components, LoadContext, hierarchical skills, PRD format, voice
announcements, context compaction). Algorithm version bumped to v3.6.0.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Releases/v4.0.1/.claude/settings.json

@@ -680,7 +680,7 @@
   "spinnerTipsOverride": {
     "excludeDefault": true,
     "tips": [
-      "77 skills loaded. Type / to see them all.",
+      "21 skill categories, 338 workflows. Type / to see them all.",
       "/council spawns a multi-agent debate on any topic.",
       "/redteam runs 32 adversarial agents against your idea.",
       "/extract-wisdom adapts its sections to match the content.",
@@ -742,25 +742,25 @@
       "Every ISC criterion must be 8-12 words, state not action, binary testable.",
       "Anti-criteria capture what must NOT happen. Prefix with ISC-A.",
       "The Quality Gate checks count, length, state, testability, and anti-criteria.",
-      "Effort levels: Instant < Fast < Standard < Extended < Advanced < Deep < Comprehensive.",
+      "Effort levels: Standard < Extended < Advanced < Deep < Comprehensive. Five tiers.",
       "Standard is the default effort level (~2 min). Only escalate when demanded.",
-      "OBSERVE is thinking-only. No tool calls except TaskCreate and context recovery.",
+      "OBSERVE creates the PRD, writes ISC criteria, and does context recovery (Grep/Glob/Read ≤34s).",
       "Context recovery uses Grep/Glob/Read directly — never agents. 34-second hard gate.",
-      "Plan mode is the ISC construction workshop. Triggers at Extended+ effort.",
+      "Plan mode (EnterPlanMode) triggers at Advanced+ effort level.",
       "PRDs persist ISC to disk. They survive across sessions. PRD on disk wins conflicts.",
       "Loop mode re-runs the Algorithm against failing PRD criteria until all pass.",
       "The Algorithm CLI: bun algorithm.ts -m loop -p PRD-file.md -n 128",
       "Parallel loop mode: bun algorithm.ts -m loop -p PRD-file.md -n 20 -a 8",
-      "15 specialized agents: Algorithm, Engineer, Architect, Designer, and more.",
+      "14 specialized agents: Algorithm, Engineer, Architect, Designer, and more.",
       "Researcher agents: Claude, Gemini, Grok, Perplexity, Codex — each with unique style.",
       "BrowserAgent uses Playwright CLI for headless browser automation.",
       "UIReviewer validates user stories — one agent per story, parallel execution.",
       "QATester is mandatory Gate 4 before claiming any web implementation is complete.",
-      "The Intern agent is a 176-IQ genius generalist for complex multi-faceted problems.",
+      "Agent templates live in agents/ — one .md file per personality with voice and behavioral rules.",
       "Agent teams need CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 (already enabled).",
       "Say 'create an agent team' to trigger multi-agent collaboration.",
       "Custom agents use subagent_type='general-purpose', never built-in type names.",
-      "20 hooks run across the session lifecycle — security, tracking, learning, naming.",
+      "21 hooks run across the session lifecycle — security, tracking, learning, naming.",
       "SecurityValidator.hook.ts guards Bash, Edit, Write, and Read operations.",
       "RatingCapture.hook.ts captures 1-10 ratings on UserPromptSubmit.",
       "SessionAutoName.hook.ts generates four-word session names automatically.",
@@ -791,35 +791,35 @@
       "First principles over bolt-on solutions. Find the root cause.",
       "The Algorithm concept: Current State to Ideal State via verifiable hill-climbing.",
       "Euphoric Surprise is the goal. Chase 9-10 ratings on every response.",
-      "ISC count scales with complexity: 4-16 simple, 17-32 medium, 33-99 large, 100+ massive.",
-      "Capability audit scans 25 capabilities every time. Format scales by effort level.",
-      "Skills, agents, thinking, collaboration, execution, verification — 6 capability sections.",
+      "ISC count floors by effort tier: Standard 8, Extended 16, Advanced 24, Deep 40, Comprehensive 64.",
+      "Capability selection consults the skill listing in the system prompt. Invoke every selected capability via tool.",
+      "Selecting a capability creates a binding commitment to invoke it via Skill or Agent tool call.",
       "The Research skill is mandatory for ALL research. Never spawn ad-hoc research agents.",
       "Council and RedTeam are collaboration capabilities — multiple perspectives on one problem.",
       "Parallelization uses run_in_background: true for concurrent agent work.",
       "Git branching via worktrees enables isolated experiments.",
       "Test-first: write or run verification alongside artifacts, not after.",
       "No silent stalls. Every command completes quickly or runs in background.",
       "5-second timeout on infrastructure commands. If hung, kill and retry.",
-      "Use existing management tools (manage.sh, CLI tools). Don't improvise Bash chains.",
+      "Use PAI Tools (bun PAI/Tools/*.ts) and CLI tools. Don't improvise Bash chains.",
       "Phase discipline: 7 separate phases. BUILD is not EXECUTE. Never merge or skip.",
       "spinnerVerbs in settings.json customizes the animated verb shown while working.",
       "The meaning of PAI: magnifying human capabilities through general problem-solving.",
-      "PAI v4.0.0 with Algorithm v3.5.0 — the latest system architecture.",
+      "PAI v4.0.1 with Algorithm v3.6.0 — the latest system architecture.",
       "The counts section in settings.json tracks skills, signals, files, and sessions.",
-      "contextFiles in settings.json controls which files auto-load into every session.",
-      "Constraint Extraction (v1.3.0) prevents specific numbers from becoming vague ISC.",
-      "Verification Rehearsal simulates violations to test if VERIFY would catch them.",
-      "The Verify Completion Gate ensures every PASS claim has a matching TaskUpdate call.",
+      "v4.0: contextFiles is empty. Core context in CLAUDE.md (native). Dynamic context via LoadContext.hook.ts.",
+      "The Splitting Test checks every ISC: and/with test, independent failure, scope words, domain boundaries.",
+      "VERIFY checks every ISC criterion AND confirms every selected capability was actually invoked via tool.",
+      "As each ISC criterion passes, immediately edit the PRD: change [ ] to [x] and update progress count.",
       "No rubber-stamp verification. Every PASS needs specific evidence, not just 'looks fine'.",
       "Wisdom Frame domains: development, deployment, communication, system-architecture.",
-      "Algorithm Reflection asks 3 questions: Self, Algorithm, and AI — what would each do differently?",
-      "ISC confidence tags: [E] Explicit, [I] Inferred, [R] Reverse-engineered.",
-      "ISC priority tags: [CRITICAL] from constraints, [IMPORTANT] inferred, [NICE] ideal state.",
-      "Loop mode effort decay: iterations 4+ drop to Standard, iterations 8+ drop to Fast.",
-      "PRD status progression: DRAFT → CRITERIA_DEFINED → PLANNED → IN_PROGRESS → VERIFYING → COMPLETE.",
+      "Algorithm LEARN phase asks 4 questions: Self, Algorithm, Capabilities, and AI — what would each do differently?",
+      "OBSERVE reverse-engineers: explicit wants, implied wants, explicit not-wanted, implied not-wanted, speed.",
+      "Anti-criteria (ISC-A prefix) capture what must NOT happen. Every PRD should have at least one.",
+      "Loop mode effort decay: later iterations automatically drop processing intensity to prevent thrashing.",
+      "PRD phase progression: observe → think → plan → build → execute → verify → learn → complete.",
       "The 2-second rule: if Grep/Glob/Read can do it in 2s, never spawn an agent.",
-      "Don't Create Too Few ISC. Extended+ needs 16-64 for small, 128-2048 for large projects.",
+      "ISC Count Gate is mandatory. Cannot exit OBSERVE with fewer ISC than the effort tier floor.",
       "Algorithm loop mode: autonomous iteration against PRD criteria until all pass.",
       "Loop mode spawns claude -p sessions focused on failing criteria each iteration.",
       "Parallel loop: -a 4 assigns one criterion per agent for simultaneous work.",
@@ -868,7 +868,20 @@
       "For architecture decisions, use FirstPrinciples + Council + RedTeam in sequence.",
       "For product strategy, use Science + WorldThreatModelHarness + Council for full analysis.",
       "For creative work, use BeCreative + IterativeDepth to maximize diversity of options.",
-      "The capability audit's ISC Improvement step checks which thinking skills would sharpen criteria."
+      "The capability audit's ISC Improvement step checks which thinking skills would sharpen criteria.",
+      "BuildCLAUDE.ts composes CLAUDE.md from Components/ at session start — modular, versionable.",
+      "LoadContext.hook.ts injects dynamic context (relationship, learning, work) at session start.",
+      "v4.0 hierarchical skills: 12 public categories replace 37 flat directories. Faster routing.",
+      "Private skills prefixed with _ (e.g., _COMMS, _FEED) are excluded from public PAI releases.",
+      "DocIntegrity.hook.ts and DocRebuild.hook.ts keep documentation cross-references current on Stop.",
+      "SkillGuard.hook.ts validates skill invocations. AgentExecutionGuard.hook.ts controls agent spawning.",
+      "v4.0 eliminated packs. The release IS the .claude/ directory — install by copying directly.",
+      "PRD frontmatter has 8 fields: task, slug, effort, phase, progress, mode, started, updated.",
+      "PRD body has 4 sections: Context, Criteria (ISC checkboxes), Decisions, Verification.",
+      "Voice announcements at Algorithm entry and every phase transition via localhost:8888/notify.",
+      "Only the primary agent makes voice curls. Subagents and teammates skip voice entirely.",
+      "PRDSync.hook.ts fires on Write/Edit of PRD.md and syncs frontmatter to work.json for the dashboard.",
+      "Context compaction at phase boundaries prevents context rot — the #1 cause of late-phase failures."
     ]
   },
   "plansDirectory": "Plans/",
@@ -953,8 +966,8 @@
   },
   "pai": {
     "repoUrl": "github.com/danielmiessler/PAI",
-    "version": "4.0.0",
-    "algorithmVersion": "3.5.0"
+    "version": "4.0.1",
+    "algorithmVersion": "3.6.0"
   },
   "preferences": {
     "temperatureUnit": "fahrenheit"

Releases/v4.0.1/README.md

@@ -8,7 +8,7 @@
 [![Categories](https://img.shields.io/badge/Categories-13-3B82F6?style=flat)](../../skills/)
 [![Hooks](https://img.shields.io/badge/Hooks-21-F97316?style=flat)](../../hooks/)
 [![Workflows](https://img.shields.io/badge/Workflows-180-8B5CF6?style=flat)](../../skills/)
-[![Algorithm](https://img.shields.io/badge/Algorithm-v3.5.0-D97706?style=flat)](../../PAI/Algorithm/)
+[![Algorithm](https://img.shields.io/badge/Algorithm-v3.6.0-D97706?style=flat)](../../PAI/Algorithm/)
 
 </div>