LifeOS

Generates static visual content across 20+ formats via Flux, Nano Banana Pro (Gemini 3 Pro), and GPT-Image-1. Covers blog header illustrations, editorial art, Mermaid flowcharts, technical architecture diagrams, D3.js dashboards, taxonomies, timelines, 2x2 framework matrices, comparisons, annotated screenshots, recipe cards, aphorism/quote cards, conceptual maps, stat cards, comic panels, YouTube thumbnails, PAI pack icons, and brand-logo wallpapers. Named workflows: Essay, D3Dashboards, Visualize, Mermaid, TechnicalDiagrams, Taxonomies, Timelines, Frameworks, Comparisons, AnnotatedScreenshots, RecipeCards, Aphorisms, Maps, Stats, Comics, YouTubeThumbnailChecklist, AdHocYouTubeThumbnail, CreatePAIPackIcon, LogoWallpaper, RemoveBackground. SKILLCUSTOMIZATIONS loads PREFERENCES.md, CharacterSpecs.md, and SceneConstruction.md. --remove-bg flag produces transparent-background PNG (can produce black backgrounds — verify visually). Up to 14 reference images per request (5 human, 6 object Gemini API limit). Output s

Comprehensive AI agent evaluation framework with three grader types (code-based: deterministic/fast; model-based: nuanced/LLM rubric; human: gold standard) and pass@k / pass^k scoring. Evaluates agent transcripts, tool-call sequences, and multi-turn conversations — not just single outputs. Supports capability evals (~70% pass target) and regression evals (~99% pass target). Workflows: RunEval, CompareModels, ComparePrompts, CreateJudge, CreateUseCase, RunScenario, CreateScenario, ViewResults. Integrates with THE ALGORITHM ISC rows for automated verification. Domain patterns pre-configured for coding, conversational, research, and computer-use agent types in Data/DomainPatterns.yaml. Tools: AlgorithmBridge.ts (ISC integration), FailureToTask.ts (failures → tasks), SuiteManager.ts (create/graduate/saturation-check), ScenarioRunner.ts (multi-turn simulated-user), TranscriptCapture.ts, PAIAgentAdapter.ts (wraps Inference.ts), ScenarioToTranscript.ts. Code-based graders: string_match, regex_match, binary_tests, st

Meta-prompting standard library — the PAI system for generating, optimizing, and composing prompts programmatically. Owns three pillars: Standards (Anthropic Claude 4.x best practices, context engineering principles, 1,500+ paper synthesis, Fabric pattern system, markdown-first / no-XML-tags); Templates (Handlebars-based — Briefing.hbs, Structure.hbs, Gate.hbs, DynamicAgent.hbs, and eval-specific templates Judge.hbs, Rubric.hbs, TestCase.hbs, Comparison.hbs, Report.hbs used by Agents and Evals skills); and Tools (RenderTemplate.ts for CLI/TypeScript rendering with data-content separation). Philosophy: prompts that write prompts — structure is code, content is data. Delivered 65% token reduction across PAI (53K → 18K tokens) via template extraction. Output is always a prompt to be used elsewhere, not final content. Reference files: Standards.md (complete prompt engineering guide), Tools/RenderTemplate.ts (rendering implementation). NOT FOR generating final content or answers — this skill produces prompts only

Compose CUSTOM agents from Base Traits + Voice + Specialization, and manage predefined functional TEAMS. Traits combine expertise (security, technical, research), personality (skeptical, analytical, enthusiastic), and approach (thorough, rapid, systematic). ComposeAgent.ts merges base + user config, outputs unique prompt + ElevenLabs voice + prosody. Predefined teams: engineering, architecture, marketing, design, security, research, content, strategy — each YAML-configured with roles, tensions, and specialist members. Observer team variant: read-only oversight agents that vote continue/halt/escalate against the tool-activity audit log (high-blast-radius or unattended runs only). USE WHEN create custom agents, spin up agents, specialized agents, agent personalities, available traits, list traits, agent voices, compose agent, spawn parallel agents, launch agents, engineering team, architecture team, marketing team, design team, security team, research team, content team, strategy team, get the team on this, obs

3-pass scope oscillation that holds a question constant while shifting the scope envelope — narrow/tactical, wide/strategic, then synthesis — to surface design tensions invisible at any single zoom level. Requires two distinct inputs: the tactical target (what you're building) and strategic context (the larger system it serves). Pass 1 captures the component's own internal logic. Pass 2 reveals what the system needs it to be. Pass 3 finds where those views diverge — that delta is the output. Produces: design tensions, scope recommendations, and coherence assessments. Single workflow: Workflows/Oscillate.md. BPE-fragile — quarterly test recommended to verify smarter models don't naturally oscillate scope without prompting. Best integration point: Algorithm OBSERVE phase (before ISC) or THINK phase (before approach commitment). NOT a lens rotation (that's IterativeDepth) and NOT idea generation (that's BeCreative). NOT FOR deep incident causal chains (use RootCauseAnalysis) or assumption decomposition (use Firs

Manages a curated aphorism collection with full CRUD — content-based matching, themed search, thinker research, and database maintenance. Organizes quotes by author, theme, context, and newsletter usage history to prevent repetition. Four workflows: FindAphorism (analyze newsletter content, match themes, return 3-5 ranked recommendations with rationale), AddAphorism (parse quote + author, extract themes, validate uniqueness, update theme index), ResearchThinker (deep research on philosopher, add sourced quotes to database), SearchAphorisms (search by theme, keyword, or author). Database at ~/.claude/skills/aphorisms/Database/aphorisms.md — stores full quote text, author attribution, theme tags, context/background, source reference, and usage history per entry. Theme index supports 12+ categories: Work Ethic, Resilience, Learning, Stoicism, Risk, Wisdom, Truth-seeking, Excellence, Curiosity, Freedom, Rationality, Clarity. Supported thinkers: Hitchens, Feynman, Deutsch, Sam Harris, Spinoza, plus any requested a

Scrape social media platforms, business data, and e-commerce via Apify actors — Instagram profiles/posts/hashtags/comments, LinkedIn profiles/jobs/posts, TikTok profiles/hashtags/videos/comments, YouTube channels/search/comments, Facebook posts/groups/comments, Google Maps business search with contact/review/image extraction, Amazon products/reviews/pricing, and general-purpose multi-page web crawling with custom pageFunction extraction logic. File-based TypeScript wrappers (scrapeInstagramProfile, searchGoogleMaps, scrapeAmazonProduct, scrapeWebsite, etc.) filter and transform data in code before returning to model context, achieving 95-99% token savings over direct MCP protocol. Parallel multi-platform queries via Promise.all for social listening dashboards. Lead enrichment pipeline: Google Maps → qualified filter → optional LinkedIn enrichment. Competitive analysis across Instagram, YouTube, and TikTok simultaneously. USE WHEN scrape Instagram, scrape LinkedIn, scrape TikTok, scrape YouTube, scrape Faceboo

Generates static visual content across 20+ formats via Flux, Nano Banana Pro (Gemini 3 Pro), and GPT-Image-1. Covers blog header illustrations, editorial art, Mermaid flowcharts, technical architecture diagrams, D3.js dashboards, taxonomies, timelines, 2x2 framework matrices, comparisons, annotated screenshots, recipe cards, aphorism/quote cards, conceptual maps, stat cards, comic panels, YouTube thumbnails, PAI pack icons, and brand-logo wallpapers. Named workflows: Essay, D3Dashboards, Visualize, Mermaid, TechnicalDiagrams, Taxonomies, Timelines, Frameworks, Comparisons, AnnotatedScreenshots, RecipeCards, Aphorisms, Maps, Stats, Comics, YouTubeThumbnailChecklist, AdHocYouTubeThumbnail, CreatePAIPackIcon, LogoWallpaper, RemoveBackground. SKILLCUSTOMIZATIONS loads PREFERENCES.md, CharacterSpecs.md, and SceneConstruction.md. --remove-bg flag produces transparent-background PNG (can produce black backgrounds — verify visually). Up to 14 reference images per request (5 human, 6 object Gemini API limit). Output s

Search and retrieve arXiv academic papers by topic, category, or paper ID — with AlphaXiv-enriched AI-generated overviews. Uses arXiv Atom API (no auth) for discovery and search across cs.AI, cs.LG, cs.CL (NLP/LLMs), cs.CR (security), cs.MA (multi-agent), cs.SE, and cs.IR. Supports title (ti:), abstract (abs:), author (au:), and category (cat:) search fields with boolean operators (AND, OR, ANDNOT); sorts by lastUpdatedDate or relevance; paginates up to 2,000 results per call with 3s rate limit between calls. AlphaXiv enrichment fetches markdown summaries from alphaxiv.org/overview/{ID}.md; full text from alphaxiv.org/abs/{ID}.md as fallback; 404 means summary not yet generated. Workflows: Latest (new papers by category), Search (topic/keyword search), Paper (single paper deep-dive by ID or URL). API returns Atom XML — parse with text processing, not jq. HTTPS required with -L flag; check published date not lastUpdatedDate for truly new submissions. Output: paper title, authors, abstract, AlphaXiv summary, an

AI-powered audio and video editing pipeline: Whisper word-level transcription (insanely-fast-whisper on MPS) → Claude segment classification (KEEP / CUT_FILLER / CUT_FALSE_START / CUT_STUTTER / CUT_DEAD_AIR / CUT_EDIT_MARKER) → ffmpeg execution with 40ms qsin crossfades and room-tone gap fill → optional Cleanvoice API cloud polish for mouth sounds and loudness normalization. Distinguishes rhetorical pauses from accidental ones. Breaths attenuated to 50% volume (not removed). Preview mode (--preview flag) shows proposed cuts without modifying audio. Aggressive mode (--aggressive flag) applies tighter filler detection thresholds. Polish step (--polish flag) uploads to Cleanvoice API for mouth sound removal and loudness normalization — confirm before cloud upload of sensitive content. Pipeline tools: Transcribe.ts, Analyze.ts, Edit.ts, Polish.ts, Pipeline.ts. Single workflow: Clean.md. Requires ANTHROPIC_API_KEY; CLEANVOICE_API_KEY optional for polish step. USE WHEN: clean audio, edit audio, remove filler words,

Divergent ideation and corpus expansion using Verbalized Sampling + extended thinking. Single-shot mode generates 5 internally diverse candidates (p<0.10 each) and surfaces the strongest. Multi-turn mode expands a small seed corpus (5-20 examples) into a diverse N-example dataset for evals, training, or test sets. Research-backed: Zhang et al. 2025 (arXiv:2510.01171) — 1.6-2.1x diversity increase on creative writing, 25.7% quality improvement, and synthetic-data downstream accuracy lift 30.6% → 37.5% on math benchmarks. Seven workflows: StandardCreativity, MaximumCreativity, IdeaGeneration, TreeOfThoughts, DomainSpecific, TechnicalCreativityGemini3 for algorithmic/architecture work, and SyntheticDataExpansion for VS-Multi corpus growth. Single-shot output is one best response, not a ranked list; SyntheticDataExpansion writes a JSONL corpus to MEMORY/WORK/{slug}/synthetic-data/. Integrates with XPost, LinkedInPost, Blogging (creative angles), Art (diverse image prompt ideas), Business (offer frameworks), Resea

Audits any AI instruction set for over-prompting using the core test: would a smarter model make this rule unnecessary? Applies Five Questions to every rule — Does Claude already do this? Contradiction? Redundant? One-off fix? Vague? — then classifies each as CUT / RESOLVE / MERGE / EVALUATE / SHARPEN / MOVE / KEEP. Two workflows: Audit (full system — reads all force-loaded files from settings.json, reports token savings estimate) and QuickCheck (single file, fast keep/cut/sharpen verdict). Outputs categorized report with estimated line and token savings. Core principle: less scaffolding = better output — every unnecessary rule competes for attention and degrades the rules that matter. Anti-fragile rules to KEEP: verification harnesses, ISC, data pipelines, specific DO/DON'T examples, tool preferences, routing rules. Fragile rules to CUT: CoT orchestrators, format parsers, retry cascades, numeric personality scales, abstract value statements. Requires loadAtStartup and postCompactRestore.fullFiles to stay in

4-tier progressive scraping with automatic escalation: Tier 1 WebFetch (fast, built-in), Tier 2 curl with Chrome headers (basic bot bypass), Tier 3 agent-browser (headless JavaScript rendering via Rust CLI daemon), Tier 4 Bright Data MCP proxy (CAPTCHA, advanced bot detection, residential proxies). Two workflows: FourTierScrape for single URLs, Crawl for multi-page site mapping (light crawl via scrape_batch loop up to 50 pages, or full crawl via Bright Data Crawl API). Always starts at Tier 1 and escalates only when blocked — Tier 4 has usage costs. Outputs URL content in markdown format. USE WHEN Bright Data, scrape URL, web scraping, bot detection, crawl site, CAPTCHA, can't access, site blocking, extract page content, scrape whole site, spider domain, convert URL to markdown, getting blocked. NOT FOR headless batch automation without scraping need (use Browser). NOT FOR simple public content (use WebFetch directly). NOT FOR real-browser bot bypass where staying logged in and zero CDP fingerprint matter (us

Headless browser automation via agent-browser — Rust CLI daemon with persistent auth profiles for fast, scriptable, parallel browser work. Supports batch commands, network interception, device emulation, per-site profile auth (one-time headed login, headless forever after), and parallel isolated sessions via --session. Workflows: ReviewStories (fan out YAML user stories to parallel UIReviewers), Automate (load/run parameterized recipe templates), Update. Delegates to general-purpose agents with agent-browser instructions for background parallel scraping. Falls back to Interceptor if site has bot detection. USE WHEN headless browser, batch scrape, fast screenshot, dev server test, parallel browser, background automation, extract data, review stories, automate recipe, batch screenshots, scrape multiple pages in parallel. NOT FOR deploy verification or UI confirmation with real Chrome (use Interceptor). NOT FOR simple single-URL fetching (use WebFetch). NOT FOR CAPTCHA or bot-detection bypass (use BrightData or

2-phase context search across the PAI session registry, work directories, and ISAs for instant cold-start recovery. Phase 1 (always): parallel scan of work.json session registry, session-names.json, MEMORY/WORK/ directory names, and ISA title grep — returns 5 most recent matches per source, loads top-3 ISA summaries (first 10 lines only). Phase 2 (only if Phase 1 returns fewer than 3 matches): PAI git history + current project git history. Output: compact context block under 40 lines with session slugs, phases, progress, and ISA paths. Reads MEMORY/STATE/work.json (task, phase, progress, effort) and MEMORY/STATE/session-names.json (sessionId, name); uses fd for directory scan with --max-depth 1 and Grep for ISA title matches in files_with_matches mode. Git history searched via `git log --oneline --all --grep` on both PAI and current project repo. Output: session slugs (newest first), ISA summaries (first 10 lines), commit hashes if Phase 2 ran. Standalone mode: present results then ask what to do. Paired mode

Multi-agent collaborative debate that produces visible round-by-round transcripts with genuine intellectual friction. All council members are custom-composed via ComposeAgent (Agents skill) with domain expertise, unique voice, and personality tailored to the specific topic — never built-in generic types. ComposeAgent invoked as: bun run ~/.claude/skills/Agents/Tools/ComposeAgent.ts. Two workflows: DEBATE (3 rounds, full transcript + synthesis, parallel execution within rounds, 40-90 seconds total) and QUICK (1 round, fast perspective check). Context files: CouncilMembers.md (agent composition instructions), RoundStructure.md (three-round structure and timing), OutputFormat.md (transcript format templates). Agents are designed per debate topic to create real disagreement; 4-6 well-composed agents outperform 12 generic ones. Council is collaborative-adversarial (debate to find best path); for pure adversarial attack on an idea, use RedTeam instead. NOT FOR parallel task execution across agents (use Delegation s

Generate production-ready TypeScript CLIs using a 3-tier template system: Tier 1 llcli-style manual arg parsing (zero deps, Bun + TypeScript, ~300-400 lines — 80% of cases), Tier 2 Commander.js (subcommands, nested options, auto-help — 15%), Tier 3 oclif reference only (enterprise scale — 5%). Every generated CLI includes full implementation, README + QUICKSTART docs, package.json with Bun, tsconfig strict mode, type-safe throughout, JSON output, exit code compliance. Workflows: CreateCli (from scratch), AddCommand (extend existing), UpgradeTier (migrate Tier 1 → 2). Outputs go to ~/.claude/Bin/ or ~/Projects/. Tier 1 (llcli pattern — 327 lines, zero deps) suits API clients, data transformers, simple automation — 2-10 commands, JSON output. Tier 2 (Commander.js) for 10+ commands, nested options, or plugin architecture. Tier 3 (oclif) is documentation-only reference for Heroku/Salesforce scale. Output tree: {name}.ts + package.json + tsconfig.json (strict) + .env.example + README.md + QUICKSTART.md. Quality ga

Complete PAI skill development lifecycle across two tracks. Structure track: scaffold new skills (TitleCase dirs, flat 2-level max, Workflows/ + Tools/ + References/ only), validate against canonical format, canonicalize existing skills. Effectiveness track (Anthropic methodology): TestSkill spawns with-skill vs baseline agents in parallel and compares outputs, ImproveSkill diagnoses root causes and rewrites instructions with reasoning over rigid constraints, OptimizeDescription generates 20 should/shouldn't-trigger test queries and rewrites for accuracy. Guides from Thariq Shihipar (Mar 2026): Gotchas section mandatory, BPE check before finalizing, progressive disclosure (frontmatter → SKILL.md body → reference files), on-demand hooks. USE WHEN create skill, new skill, validate skill, test skill, improve skill, optimize description, skill not triggering, skill overtriggering, canonicalize, scaffold skill, skill quality. NOT FOR TypeScript CLI generation (use CreateCLI).

Manage the public daemon profile — a living digital representation of what you're working on, thinking about, reading, and building. DaemonAggregator.ts reads PAI sources (TELOS missions/goals/books/wisdom, KNOWLEDGE/Ideas titles, PROJECTS.md, MEMORY/WORK themes, PRINCIPAL_IDENTITY bio) and writes to daemon-data.json. SecurityFilter.ts applies deterministic pattern-matching (NOT LLM judgment) to strip names, paths, credentials, and internal refs. Structurally excludes CONTACTS, FINANCES, HEALTH, OUR_STORY, OPINIONS. deploy.sh builds the VitePress static site and deploys to Cloudflare Pages. Two-repo pattern: public framework (danielmiessler/Daemon, forkable) + private content (daemon-dm). Workflows: UpdateDaemon, ReadDaemon, PreviewDaemon, DeployDaemon. USE WHEN daemon, update daemon, daemon profile, deploy daemon, preview daemon, read daemon, check daemon, daemon status, public profile, digital presence. NOT FOR internal PAI system management (use _PAI).

Parallelize work via six patterns: built-in agents (Engineer/Architect/Algorithm/Explore/Plan via Task), worktree-isolated agents (conflict-free parallel file edits), background agents (run_in_background: true, non-blocking), custom agents (ComposeAgent via Agents skill → Task general-purpose), agent teams (TeamCreate + TaskCreate + SendMessage for multi-turn peer coordination), and parallel task dispatch (N identical operations). Two-tier delegation: lightweight (haiku, max_turns=3, one-shot extraction/classification) vs full (multi-step, tool use, iteration). Decision rule — agents need to talk to each other or share state → Teams; independent one-shot work → Subagents. Auto-invoked by Algorithm when 3+ independent workstreams exist at Extended+ effort. USE WHEN 3+ workstreams, parallel execution, agent specialization, agent team, swarm, spawn agents, create team, fan out, divide and conquer, multi-agent, coordinate agents. NOT FOR single-agent custom personality composition (use Agents skill).

Comprehensive AI agent evaluation framework with three grader types (code-based: deterministic/fast; model-based: nuanced/LLM rubric; human: gold standard) and pass@k / pass^k scoring. Evaluates agent transcripts, tool-call sequences, and multi-turn conversations — not just single outputs. Supports capability evals (~70% pass target) and regression evals (~99% pass target). Workflows: RunEval, CompareModels, ComparePrompts, CreateJudge, CreateUseCase, RunScenario, CreateScenario, ViewResults. Integrates with THE ALGORITHM ISC rows for automated verification. Domain patterns pre-configured for coding, conversational, research, and computer-use agent types in Data/DomainPatterns.yaml. Tools: AlgorithmBridge.ts (ISC integration), FailureToTask.ts (failures → tasks), SuiteManager.ts (create/graduate/saturation-check), ScenarioRunner.ts (multi-turn simulated-user), TranscriptCapture.ts, PAIAgentAdapter.ts (wraps Inference.ts), ScenarioToTranscript.ts. Code-based graders: string_match, regex_match, binary_tests, st

Content-adaptive wisdom extraction that reads the content first, detects what wisdom domains are actually present, then builds custom sections around what it finds — instead of forcing static headers every time. A security talk gets 'Threat Model Insights' and 'Defense Strategies'; a business podcast gets 'Contrarian Business Takes' and 'Money Philosophy'. Five depth levels: Instant (1 section), Fast (3 sections), Basic (3+takeaway), Full (5-12 sections, default), Comprehensive (10-15+themes). Voice follows the user's conversational style — bullets read like telling a friend what you just watched, not a press release. Output always includes dynamic sections, One-Sentence Takeaway, 'If You Only Have 2 Minutes', and References. Spicy/contrarian takes are mandatory inclusions, never softened. YouTube content extracted via `fabric -y URL` before extraction; article content fetched via WebFetch. Output: markdown with dynamic section headers, closing sections vary by depth level, References & Rabbit Holes, optional

Execute any of 240+ specialized prompt patterns natively (no CLI required for most) across categories: Extraction (30+), Summarization (20+), Analysis (35+), Creation (50+), Improvement (10+), Security (15), Rating (8). Common patterns: extract_wisdom, summarize, create_5_sentence_summary, create_threat_model, create_stride_threat_model, analyze_claims, improve_writing, review_code, extract_main_idea, analyze_malware, create_sigma_rules, create_mermaid_visualization, youtube_summary, judge_output, rate_ai_response. Native execution reads Patterns/{name}/system.md and applies directly. Fabric CLI used only for YouTube transcript extraction (-y URL) and fallback URL fetching (-u URL). Patterns auto-synced from upstream Fabric repo via UpdatePatterns workflow. Workflows: ExecutePattern, UpdatePatterns. USE WHEN fabric, fabric pattern, run fabric, sync fabric, update patterns, use extract_wisdom, threat model, summarize, analyze claims, improve writing, review code, create prd, rate content, create diagram, merma

Physics-based reasoning framework (Musk/Elon methodology) that deconstructs problems to irreducible fundamental truths rather than reasoning by analogy. Three-step structure: DECONSTRUCT (break to constituent parts and actual values), CHALLENGE (classify every element as hard constraint / soft constraint / unvalidated assumption — only physics is truly immutable), RECONSTRUCT (build optimal solution from fundamentals alone, ignoring inherited form). Outputs: constituent-parts breakdown, constraint classification table, and reconstructed solution with key insight. Three workflows: Deconstruct.md, Challenge.md, Reconstruct.md. Integrates with RedTeam (attack assumptions before deploying adversarial agents), Security (decompose threat model), Architecture (challenge design constraints), and Pentesters (decompose assumed security boundaries). Other skills invoke via: Challenge on all stated constraints → classify as hard/soft/assumption. Cross-domain synthesis: solutions from unrelated fields often apply once the

Evolutionary ideation engine — loop-controlled multi-cycle idea generation through 9 phases (CONSUME, DREAM at noise=0.9, DAYDREAM at noise=0.5, CONTEMPLATE at noise=0.1, STEAL cross-domain borrowing, MATE recombination via Fisher-Yates shuffle, TEST fitness scoring, EVOLVE selection, META-LEARN Lamarckian strategy adjustment). Loop Controller drives adaptive continue/pivot/stop logic with mid-cycle quality checkpoints; strategies evolve across cycles based on what worked. Produces ranked novel solution candidates with full provenance and fitness landscape. Six workflows: FullCycle (all 9 phases adaptive — default), QuickCycle (compressed CONSUME+STEAL+MATE+TEST single cycle), Dream (DREAM phase only), Steal (cross-domain transfer only), Mate (recombination only), Test (fitness evaluation only). Integrates IterativeDepth in CONTEMPLATE, RedTeam in TEST, Council optionally in MATE. NOT FOR quick single-pass brainstorming (use BeCreative). USE WHEN ideate, id8, novel ideas, generate ideas, ideation engine, evol

Real Chrome browser automation via Interceptor extension — controls the actual browser from inside (zero CDP fingerprint, passes all major bot detection checks including BrowserScan, Pixelscan, CreepJS, Fingerprint.com). Stays logged in, uses your real sessions. Compound commands (open, read, act, inspect) collapse multi-step flows into single calls. Unique capabilities: monitor/replay system (record user actions → export replayable plan scripts for regression), network log (auto-captures all fetch/XHR), scene graph for rich editors (Google Docs, Canva, Slides). Workflows: VerifyDeploy, Reproduce (open affected page BEFORE code analysis — mandatory per rules), RecordFlow, ReplayFlow, TestForm, Update. MANDATORY for all visual verification — never use agent-browser for deploy confirmation. USE WHEN verify deploy, confirm UI, check page, screenshot verification, interceptor, debug web, troubleshoot, visual check, authenticated page, bot detection bypass, agent-browser failing, reproduce bug, record flow, replay

Runs a phased conversational interview across all PAI context files using InterviewScan.ts, which orders targets by PHASE and assigns conversation mode per file. Phase 1 (foundational TELOS) always runs first regardless of completeness: MISSION, GOALS, PROBLEMS, STRATEGIES, CHALLENGES, NARRATIVES, SPARKS, BELIEFS, WISDOM, MODELS, FRAMES in leverage order. Phase 2: IDEAL_STATE (HEALTH, MONEY, FREEDOM, RELATIONSHIPS, CREATIVE) in Fill mode. Phase 3: preferences (BOOKS, AUTHORS, BANDS, MOVIES, RESTAURANTS, FOOD_PREFERENCES, LEARNING, MEETUPS, CIVIC) in mixed mode. Phase 4: light touch on CURRENT_STATE/SNAPSHOT and PRINCIPAL_IDENTITY. Phase 9 (RHYTHMS) deferred. Review mode (≥80%) reads file then asks targeted questions one at a time — still accurate, outdated, missing, sharpen? Fill mode (<80%) walks scanner prompts one at a time. The principal answers in natural language; the DA formats into file structure. Voice confirms on actual changes only. Stop signals respected immediately. Target vs. north-star type con

Owns the Ideal State Artifact — the universal primitive that holds the articulated ideal state of any thing (project, task, application, library, infrastructure, work session) as a hard-to-vary explanation. The ISA is a single document that articulates the ideal state, drives the build, verifies the build, and records the evolution of understanding. Five workflows: Scaffold (generate a fresh ISA from a prompt at a specified effort tier), Interview (adaptive question-and-answer to fill in or deepen sections), CheckCompleteness (score an existing ISA against the tier completeness gate and report gaps), Reconcile (deterministic merge of an ephemeral feature-file excerpt back into the master ISA, keyed on stable ISC IDs), Seed (bootstrap a draft project ISA from an existing repository's README, code structure, and recent commits). Examples directory contains canonical-isa.md (the showpiece reference, fully populated across all twelve sections), e1-minimal.md (90-second task — Goal + Criteria only), e3-project.md

Structured multi-angle exploration that runs 2-8 sequential passes through the same problem, each from a systematically different scientific lens, to surface requirements and edge cases invisible from any single angle. Grounded in 20 established techniques across cognitive science (Hermeneutic Circle, Triangulation), AI/ML (Self-Consistency, Ensemble Methods), requirements engineering (Viewpoint-Oriented RE), and design thinking (Six Thinking Hats, Causal Layered Analysis). Each pass outputs new ISC criteria; passes stop when yields repeat. Best used in the OBSERVE phase at Extended+ effort — the default question should be why NOT to use it, not why to use it. A 4-lens pass routinely discovers 30-50% more criteria than direct analysis. Single workflow: Workflows/Explore.md (supports Fast mode with 2 lenses for quick depth). Reference files: ScientificFoundation.md (research grounding for all 20 techniques), TheLenses.md (full definitions for all 8 lenses). BPE-fragile — quarterly test recommended. NOT FOR sco

Manage the PAI Knowledge Archive — a curated, typed graph of notes across four entity domains: People, Companies, Ideas, and Research. Operations: search (3-pass: lexical + frontmatter + wikilink), add (creates note with mandatory typed cross-links), harvest (KnowledgeHarvester pulls from PAI sources), develop (surfaces seedling notes for enrichment), ingest (fetch URL or file, create primary note, ripple updates to related notes), contradictions (find conflicting claims via tag-overlap pairs), graph (stats or 2-hop traversal via KnowledgeGraph.ts), retrieve (BM25-lite compressed context via MemoryRetriever.ts), mine (SessionHarvester extracts memory candidates from recent conversations). Every note ships with typed related: frontmatter links (8 relationship types: supports, contradicts, extends, part-of, instance-of, caused-by, preceded-by, related). USE WHEN knowledge, knowledge base, search knowledge, what do we know about, archive, harvest, knowledge status, develop note, add to knowledge, ingest, contrad

Iterative improvement loop — revisit and refine a target across multiple Algorithm cycles toward an ideal state. USE WHEN loop, iterate, refine, improve iteratively, multiple passes, keep improving, loop mode, revisit, rework.

Intakes existing content from external sources, classifies each chunk against the PAI destination taxonomy, and commits approved chunks with provenance. Sources: .md/.markdown/.txt, stdin, PAI TELOS/MEMORY/KNOWLEDGE dirs, CLAUDE.md/.cursorrules/OpenAI Custom Instructions, Obsidian/Notion/Apple Notes exports, journal dumps. MigrateScan.ts chunks and classifies, producing a routing table with per-target counts and confidence %. MigrateApprove.ts approval loop: --approve-all, --approve-target, --review, --dry-run. UNCLEAR never bulk-approved. Phase 6 delivers summary and /interview recommendation for sparse areas. Confidence: ≥70% auto-approve; 40-70% confirm; <40% walk-through. Destinations: TELOS (MISSION/GOALS/PROBLEMS/STRATEGIES/CHALLENGES/BELIEFS/WISDOM/MODELS/FRAMES/NARRATIVES/SPARKS), IDEAL_STATE (per-dimension explicit call), preferences (BOOKS/AUTHORS/MOVIES/BANDS/RESTAURANTS/FOOD_PREFERENCES/LEARNING/MEETUPS/CIVIC), Identity (PRINCIPAL_IDENTITY.md — always prompts), Knowledge (MEMORY/KNOWLEDGE/{Ideas,P

Autonomous optimization loop — hill-climb any target. Code with metrics, or skills/prompts/agents with LLM-as-judge eval. USE WHEN optimize, autoresearch, hill climb, improve metric, reduce latency, improve performance, benchmark optimization, bundle size, page speed, autonomous improvement loop, optimize skill, optimize prompt, improve quality, eval mode.

Generate prioritized PAI upgrade recommendations via 4 parallel threads: Thread 0 (prior-work audit — reads current Algorithm, PATTERNS.yaml, hooks, settings, recent ISAs, and KNOWLEDGE to assign Prior Status tags), Thread 1 (user context — TELOS goals, active projects, PAI system state), Thread 2 (source collection — Anthropic releases, YouTube channels, GitHub trending, custom sources), Thread 3 (internal reflections — Algorithm execution Q1/Q2 patterns). Output format: Discoveries table ranked by interestingness, then tiered Recommendations (CRITICAL/HIGH/MEDIUM/LOW) each with Prior Status (NEW/PARTIAL/DISCUSSED/REJECTED/DONE), then full Technique Details with before/after code. Every recommendation cites file:line evidence from Thread 0 — already-implemented items go to Skipped, never re-surfaced. Workflows: Upgrade, MineReflections, AlgorithmUpgrade, ResearchUpgrade, FindSources, TwitterBookmarks. USE WHEN upgrade, system upgrade, check Anthropic, new Claude features, algorithm upgrade, PAI upgrade, chec

Ethical people-finding and identity verification using 15 parallel research agents (5 types x 3 each = 45 concurrent search threads) across people search aggregators, social media, public records, and reverse lookups. Covers TruePeopleSearch, FastPeopleSearch, Spokeo, voter registration, county property records, court portals (PACER, CourtListener), professional licenses, Facebook/LinkedIn/Instagram x-ray searches, and username enumeration (Sherlock, WhatsMyName). Phone reverse lookup via CallerID, NumLookup, carrier lookup. Email reverse lookup via Epieos, Holehe, Hunter.io. Image reverse search via PimEyes, TinEye, Google/Yandex Images. Google dorking (site:linkedin.com, filetype:pdf resume) across foundation, primary, deep, and verification tiers. Produces confidence-scored results (HIGH/MEDIUM/LOW/POSSIBLE) requiring 3+ matching identifiers before acting. Workflows: FindPerson, SocialMediaSearch, PublicRecordsSearch, ReverseLookup, VerifyIdentity. Stops immediately if purpose shifts toward harassment or s

Meta-prompting standard library — the PAI system for generating, optimizing, and composing prompts programmatically. Owns three pillars: Standards (Anthropic Claude 4.x best practices, context engineering principles, 1,500+ paper synthesis, Fabric pattern system, markdown-first / no-XML-tags); Templates (Handlebars-based — Briefing.hbs, Structure.hbs, Gate.hbs, DynamicAgent.hbs, and eval-specific templates Judge.hbs, Rubric.hbs, TestCase.hbs, Comparison.hbs, Report.hbs used by Agents and Evals skills); and Tools (RenderTemplate.ts for CLI/TypeScript rendering with data-content separation). Philosophy: prompts that write prompts — structure is code, content is data. Delivered 65% token reduction across PAI (53K → 18K tokens) via template extraction. Output is always a prompt to be used elsewhere, not final content. Reference files: Standards.md (complete prompt engineering guide), Tools/RenderTemplate.ts (rendering implementation). NOT FOR generating final content or answers — this skill produces prompts only

Military-grade adversarial analysis that deploys 32 parallel expert agents (engineers, architects, pentesters, interns) to stress-test ideas, strategies, and plans — not systems or infrastructure. Two workflows: ParallelAnalysis (5-phase: decompose into 24 atomic claims → 32-agent parallel attack → synthesis → steelman → counter-argument, each 8 points) and AdversarialValidation (competing proposals synthesized into best solution). Context files: Philosophy.md (core principles, success criteria, agent types), Integration.md (how to combine with FirstPrinciples, Council, and other skills; output format). Targets arguments, not network vulnerabilities. Findings ranked by severity; goal is to strengthen, not destroy — weaknesses delivered with remediation paths. Collaborates with FirstPrinciples (decompose assumptions before attacking) and Council (Council debates to find paths; RedTeam attacks whatever survives). Also invoked internally by Ideate (TEST phase) and WorldThreatModel (horizon stress-testing). NOT F

Creates programmatic video with React via Remotion — builds compositions, sequences, and motion graphics rendered to MP4. Uses useCurrentFrame() for all animation (no CSS animations). Integrates PAI_THEME constants from Theme.ts and Art skill aesthetic preferences for visual consistency. Render command: bunx remotion render {composition-id} ~/Downloads/{name}.mp4. Output always to ~/Downloads/ for preview first. Tools: Render.ts (render, list compositions, create projects) and Theme.ts (PAI theme constants derived from Art). Reference docs: ArtIntegration.md (theme constants, color mapping), Patterns.md (code examples, presets), CriticalRules.md (what not to do), Tools/Ref-*.md (31 pattern files covering core Remotion, Lambda, ElevenLabs captions, AI pipeline). Supports ElevenLabs captions, Lambda rendering, and AI pipeline integration. Rendering is CPU-intensive — use run_in_background. Two primary workflows: ContentToAnimation (animate existing content) and GeneratedContentVideo (AI content to video / make

Comprehensive research and content extraction with 4 depth modes: Quick (1 Perplexity agent, ~10-15s), Standard (4 agents — Claude + Gemini + Grok + Perplexity, cross-checked, ~30-60s), Extensive (7 explorers + 2 independent verifiers, ~60-90s), Deep Investigation (progressive iteration with persistent MEMORY/RESEARCH/ vault, loop-compatible, ~3-60min). Every URL verified before delivery — hallucinated links are a catastrophic failure. Verification architecture: per-agent self-verification, cross-check synthesis, and independent verifier agents (Extensive/Deep). Confidence-tagged output: [HIGH] [MED] [LOW] [CONFLICT]. Additional workflows: ExtractAlpha (highest-signal insights), Retrieve (CAPTCHA/bot-blocked content), YoutubeExtraction (fabric -y), WebScraping, InterviewResearch (Tyler Cowen style), AnalyzeAiTrends, Fabric (242+ patterns), Enhance, ExtractKnowledge. USE WHEN research, do research, quick research, extensive research, deep investigation, find information, investigate, extract alpha, analyze con

Structured incident investigation grounded in Toyota Production System, Kaoru Ishikawa, James Reason's Swiss Cheese model, Dean Gano's Apollo method, and Google SRE blameless-postmortem culture. Five workflows: FiveWhys (linear/branching causal chain, single-thread incidents), Fishbone/Ishikawa (6 M's or 4 P's category mapping, multiple suspected areas), Postmortem (blameless timeline + contributing factors + action items, wraps other methods), FaultTree (AND/OR gate logic, safety-critical multi-path failures), KepnerTregoe IS/IS-NOT (distinction analysis, subtle hard-to-reproduce defects). Context files: Foundation.md (Toyoda, Ishikawa, Reason, Gano, Google SRE; canonical methods), MethodSelection.md (decision flow for workflow selection). Core axiom: proximate cause is where analysis starts, not ends. Humans are never root causes — if a human could make the mistake, the system allowed it. A cause is "root enough" when it's actionable. Also supports FMEA-style pre-launch risk inversion (what could fail befor