Prompt systems
One prompting system per deep dive and guide: a deeper-not-bigger rewrite of the ask, a verbose one-shot prompt, 5–9 step prompts that each carry their own audit, a post-build gap analysis, the questions the agent must ask itself, and the external oracles that prove the result right. Paste into Claude Code or any agent. Machine-readable at api/prompts.json.
All 64 prompt systems are free here and as JSON (api/prompts.json).
agent-architecture
Take one real agent you own and rebuild it around a clean split of context (what the model sees this turn), memory (what persists and how it is retrieved, updated and forgotten) and harness (the code that runs the loop, tools and guardrails), then prove the ch
agent-injection-ci
Build a defensive prompt-injection CI harness for YOUR OWN agent: a fixed suite of injection and 'just ask nicely' hijack cases that runs on every skill install and harness change, with a canary secret and a tool-call allowlist, failing the change if
agent-mission-control
Produces a local agent control room: Claude Code hooks emit one JSON event per agent state change onto a local queue, a small redaction step strips secrets and personal data, and a spatial front end (Bot Crossing or a 200-line Three.js scene) shows agents clus
ai-capex-credit-lens
Running this system builds a paper-only research tracker that watches the one variable neocloud GPU debt depends on: hourly GPU rental prices after contracts end, set against token price deflation. It logs daily H100 and B200 rental quotes and the cheapest per
ai-trading-agents
Build a multi-agent trading research system where specialist agents (HMM regime detector, Heston volatility surface, Cox-process event intensity, phase-space dynamics, Avellaneda-Stoikov market maker) feed a debate and a risk gate, then test whether debate add
ai-video-production
Running this system produces one finished 60-90 second video made by one person with AI tools: a Claude-written script and shot list, base footage from Runway (Gen-3 Alpha, video-to-video, non-linear keyframe canvas) and Kling (motion brush on up to 5 subjects
album-as-device
Produces a song-as-object prototype: a Raspberry Pi player that recombines the artist's own 8-12 stems by seeded rules so no two plays match, adds an optional local voice line, shows the play number on a Sharp memory LCD, and sits in a printed case with a
ambient-executive-function
Running this system builds a small always-on desk object, an ESP32 driving a 2.7 inch Sharp memory LCD in a printed stand, that shows the top three lines of NOW.md and the count of Claude Code sessions waiting for approval, plus NFC tags that log done/started/
archive-trained-taste
Build a defensible design voice from a public-domain archive nobody is mining, instead of the same borrowed toolkit everyone feeds their agent. The pieces are a scraped, license-checked corpus of 2,000 to 5,000 images from one archive; CLIP embeddings and clus
art-market-disruption
Build a working direct-to-collector practice for one artist in 30 days. The pieces are: a taste atlas extracted from their own references (the @rndyrbrts method); one interactive or generative work that a gallery wall cannot hold (MediaPipe plus TouchDesigner,
autonomous-agents-wild
Separate working autonomous-agent patterns from demo-ware by building and stress-testing one narrow agent with real guardrails: a single task, one or two tools, fast feedback, hard budget and scope caps, persistent memory, and an adversarial test suite that tr
autonomous-brand-machine
Build and test a supervised content system for one brand: an Obsidian brand vault indexed with Graphify, Claude Code generating ad and UGC-style assets in Higgsfield (Nano Banana for stills, Seedance 2.0 for video) via Playwright MCP, a review queue, and posti
avellaneda-stoikov-and-hawkes
Running this system turns the guide into a tested research stack: a verified Avellaneda-Stoikov (AS) quoting engine, a Hawkes order-flow model fitted and goodness-of-fit tested on real Hyperliquid trade data, and an event-driven paper backtest comparing a fixe
benchmark-your-brain
Produces a six-week self-experiment that runs the SkillOpt-style benchmark-gated loop on a human learner: the last 60 days of saves (a DejaViewed parquet export qualifies) become the study set, held-out retrieval quizzes become the eval set, and a coach skill
blender-mcp-tripo-pipeline
Running this system sets up Blender MCP with a hosted image-to-3D generator and drives it from Claude to turn one reference photo into a rendered Blender scene, with the integration claims checked against the code. It produces a working MCP connection, a scrip
boris-cherny-claude-md
Running this system turns Boris Cherny's CLAUDE.md pattern into a tested project file: a lean, project-specific CLAUDE.md under about 200 lines, rules that earned their place, deterministic hooks for the rules that must never break, and a small behavioura
claude-code-hooks-system
Build a tested, version-pinned hook suite for Claude Code: a destructive-Bash blocker, a secret-leak blocker on Edit/Write, auto-format on edit, git context at SessionStart, a Bash audit log, a PreCompact transcript backup, and a Stop gate that refuses 'd
claude-code-leaked-system-prompt
Turns the publicly mirrored Claude Code system prompt (Piebald-AI/claude-code-system-prompts, asgeirtj/system_prompts_leaks, Drew Breunig's 2026-04-04 assembly analysis) into a hardened, measured system prompt for YOUR OWN agent. Outputs: a version-pinned
claude-code-mastery
Configure Claude Code for one real project so it costs fewer tokens and finishes more work: a lean CLAUDE.md, primer.md, hooks, parallel worktrees, a vetted subset of the 7 repos, and output and token tooling (Caveman, codeburn). For a daily Claude Code user h
claude-code-parallel-worktrees
Stand up a working 3-5 session parallel Claude Code workflow on one real repo using git worktrees, with automatic per-worktree ports and databases, a status view of all worktrees, and a measured answer to whether it actually raised your merged-PR throughput. D
claude-harness-pattern
Produce a working, version-controlled Claude harness for one real project: a CLAUDE.md of permanent rules, a primer.md that is rewritten every session, a skills/ folder of reusable workflows, and at least one hook that self-corrects behaviour. Done means a fre
claude-mastery
Turn a bare Claude Code install into a mastered, measured setup for one real project: CLAUDE.md, skills, hooks, MCP/CLI tools, memory persistence, parallel worktrees, and cost controls - each adopted only after it beats the prior setup on a repeatable test. Do
claude-md-examples
Produce an evidence-backed CLAUDE.md for one real repository the user owns. Every line traces to a trap, command, check or ban the code cannot reveal. Irreversible rules are backed by a hook or CI check, and volatile state moves to primer.md. Done means an A/B
claude-second-brain
Build a Claude Code workspace with layered persistent memory - CLAUDE.md for rules, a lessons file fed by corrections, primer.md rewritten each session, a SessionStart hook that injects live git state, and an Obsidian vault structured as an LLM wiki with a gap
counterparty-forensics
Running this system measures the edge already being taken from you before you look for your own. It replicates the overnight-versus-intraday return split on SPY and QQQ with realistic costs and a control, measures your own broker fills against the national bes
creative-tech-stack-2026
Ship one complete creative artifact end to end as a one-person studio: an interactive 3D web experience (or a generative/motion piece) built through the full stack - Claude Code/Cursor for the build, a reusable reference library or design system for the agent,
earned-autonomy-router
Build a three-tier task router (frontier, mid-tier, and a local Qwen 27B on one 24GB GPU) plus a zero tier for plain shell commands, wired into an existing orchestrator-worker setup. Each task class earns its way down the stack only through statistically justi
eval-gated-skill-evolution
Build a two-way, eval-gated ratchet for your CLAUDE.md and skills: a held-out eval from your own history, an ablation sweep that deletes rules whose removal does not hurt, and a SkillOpt-style loop that merges only edits beating the benchmark - with a noise ch
game-theory-ai
Turn the six saved posts into a working incentive-mapping kit. That means three formal games written down and solved: the SpaceX day-one IPO trade versus the 'holdco re-rating' trade, the expired-API-domain squat that 'really thought this would
gemini-cli
Install Gemini CLI next to Claude Code and build a measured routing layer: a free commit-message and PR-description drafter, a large-context repo summarizer that feeds Claude, and a grounded-research helper, each wired through `gemini -p --output-format json`.
generative-art-code
Running this system produces a small generative-art practice in code: three seeded simulations (ant colony optimization with pheromone diffusion and evaporation, shepherd-and-flock boids with a pressure field, and a rule-based 'code garden' that grow
headless-everything
Make one legacy tool that you or a client are licensed to use agent-drivable, lawfully, and decide honestly whether wrapping it is an arbitrage or a consulting treadmill. The pieces are an agent-built command-line wrapper covering the tool's top five real
hunyuan3d-2
Run Tencent's Hunyuan3D-2 locally as a reproducible image-to-3D pipeline: reference image in, cleaned and textured GLB out, measured for VRAM, time, mesh validity, and fidelity to the input, then extended to the text-to-image-to-mesh chain and a Blender i
interactive-installations
Running this system builds a working body-reactive installation: webcam body tracking with MediaPipe (via Torin Blankensmith's TouchDesigner plugin) or a Kinect, a TouchDesigner network that turns landmarks into a particle or data visual in the style of p
l1b3rt4s-jailbreak-repo
Uses the structure of Elder Plinius's L1B3RT4S corpus (about 40 vendor-sorted .mkd files, SYSTEMPROMPTS.mkd, *SPECIAL_TOKENS.json, !SHORTCUTS.json) as input to a defensive program for YOUR OWN LLM agents. It produces a channel-by-class threat model, code-
llms-txt-standard
Ship a spec-conformant /llms.txt, an optional /llms-full.txt, and .md twins for the key pages of one real site, generated from the site's own build so they never drift, and measure whether agents actually use them. Done means the files parse with the refe
mediapipe-touchdesigner-pipeline
Build and ship one real-time, body-reactive visual: MediaPipe tracking (hands 21 / pose 33 / face 468 landmarks) piped into TouchDesigner, driving a generative visual that responds to movement, exported as a standalone or NDI stream. Done means it runs live at
melody-genealogy
Build a local, fully offline melody-genealogy engine that does by machine what Chance McCoy does by ear: cluster public-domain recordings into tune families and map how melodies travelled across centuries and languages. The pieces are a license-verified corpus
micro-bending-spoons
Produces an underwriting pipeline for buying one small cash-flowing app under 50K USD and rebuilding it with an agent stack: a scored listing table from Acquire.com and Flippa, verified financials, a repo-comprehension pack (Gitingest, DeepWiki, GitDiagram, Gi
mocap-as-reward
Running this system replaces reward-function trial and error with 'show it what you mean' on a cheap open robot. It reproduces the Microduck reward hack, captures or generates a reference motion (FreeMoCap from a webcam or HY-Motion from text), retar
n8n-mcp-and-obsidian-skills
Install czlonkowski's n8n-mcp and kepano's obsidian-skills into Claude Code safely and prove each one changes outcomes: n8n-mcp should turn natural-language requests into n8n workflows that import and execute on a disposable dev instance, and obsidia
natural-language-3d
Produce a working text-or-image-to-3D-scene pipeline: Claude Code drives Blender through the Blender MCP server, Tripo and Hunyuan3D-2 generate the meshes, and the result is a saved .blend plus a headless render that matches a written brief or a 2D reference i
offline-ai-appliances
Produces a working single-purpose offline AI appliance in two stages: first an exact rebuild of Google's Gemma Translator on a Raspberry Pi 5 to learn its deployment script, model stack and thermal limits, then a fork into a 'paperwork reader' t
one-dancer-thousand-bodies
Produces a 90-second ensemble piece in Blender where 200 to 1,000 bodies perform one dancer's captured phrases, and the choice of phrase and facing for each body is governed by an Ising or Potts herding model whose temperature the artist drops live throug
one-person-creative-studio
Run one real creative brief end to end as a single person with Claude Code as the orchestrator: brand and layout in Figma via the Figma MCP, a short video ad via Higgsfield Seedance 2.0 driven by Playwright MCP, an interactive MediaPipe plus TouchDesigner piec
onyx-self-hosted-ai
Stand up Onyx (formerly Danswer, github.com/onyx-dot-app/onyx) as a self-hosted knowledge layer over everything that is not code: project notes, MEMORY.md files, guides, GitHub repos, optionally Gmail and Drive. Claude via API does generation, local sentence-t
openscreen-screen-recorder
Install or build OpenScreen (github.com/siddharthvaddem/openscreen, MIT, Electron + React + PixiJS) on Victor's Fedora Linux Wayland desktop, prove its capture path works with PipeWire, and produce one publish-ready Claude Code workflow demo with auto-zoo
overfit-proof-autoresearch
Running this system builds a strategy autoresearch loop whose judge is designed to reject overfitting: Claude proposes strategy edits to a VectorBT backtester, every proposal is logged to a trials table, and a four-part gate (net of realistic fees, beats 95% o
plotter-loop
Running this system produces a working geometry-to-ink pipeline: Go programs on Fogleman's ln that turn OBJ meshes, a Thomas attractor, a ray-marched Mandelbulb slice and a 3-lag phase-space embedding of BTC volatility into hidden-line SVGs, a vpype recip
polymarket-99c-limit-arbitrage
Decide, with data rather than a story, whether resting 97-99c bids on near-certain Polymarket contracts earns more than the resolution-against tail it is paid to absorb. Running this produces a historical tail-calibration study, a fill and adverse-selection ba
polymarket-weather-trading
Test whether a calibrated multi-model ensemble forecast of the exact resolution station beats Polymarket's prices on daily temperature and precipitation buckets after fees and fill costs. Running this produces a point-in-time forecast archive, a probabili
predict-predictability
Produces a walk-forward study that runs the open Kronos candlestick foundation model zero-shot on two years of hourly BTC and QQQ, scores every forecast against naive baselines, and tests whether tradeability measures (Kaufman efficiency ratio, Lempel-Ziv comp
primer-md-pattern
Implement the primer.md living-handoff pattern (credited to @keshavsuki, lineage from @ykdojo's HANDOFF.md) in a real project so every Claude Code session starts from an accurate state snapshot and ends by rewriting it, without the human remembering. Runn
prompt-architecture
Turn a working LLM feature from prompt-tinkering into measured prompt architecture: an eval set over real tasks, a context pipeline (what enters the window and from where), a persisted memory layer, retrieval where it earns its place, and a self-correction loo
prompt-engineering-discipline
Running this system turns 'prompt engineering' into an engineering practice for the user's own systems: a versioned prompt library with regression evals, a system-prompt architecture review informed by published and leaked system prompts (includ
quant-physics-crossover
Turn the physics-to-markets posts (phase space reconstruction, sandpile avalanches, Ising herding, Heston stochastic volatility, Omori aftershock decay, Wasserstein drift) into tested indicators on real data, combine the survivors into a small ensemble, and ju
quant-project-ideas-list
Turn the 15-project quant list (seeded by @heidioftheeast's thread) into the one build that makes every later project trustworthy: a validation gate (Probability of Backtest Overfitting via CSCV, Probabilistic and Deflated Sharpe Ratios, Minimum Track Rec
scan-fit-print
Produces a scan-to-parametric-CAD pipeline in which a coding agent takes a mesh of a real object (from the Orbiter smartphone rig, photogrammetry, or a Raspberry Pi LiDAR handheld), fits geometric primitives to the contact surfaces, writes a CadQuery script fo
self-osint-sentinel
Build one scheduled defensive agent that runs the public-footprint toolkit against your own identities, never anyone else's, and files the remediation. The pieces are a weekly job that runs username and email enumeration on your own handles, a photo pass
selling-to-agents
Running this system turns one dataset you genuinely own into a machine-first product: a free llms.txt discovery layer, a structured JSON endpoint metered per call with x402 stablecoin micropayments, and the same data exposed as a listed MCP server, with reques
sensor-settled-contracts
Produces a settlement-source map of every open Kalshi and Polymarket weather/event contract, a two-week side-by-side log of a Raspberry Pi sensor against the exact station each contract settles on, and a paper-only trading arm for the final six hours before re
sketch-to-production
Running this system takes one hand-drawn character from a Procreate kit-bash sheet to a rigged, animated, rendered 3D shot using Jon Finger's chain (Procreate, Magnific style transfer, 3DAI Studio or Hunyuan3D-2, Move AI mocap, render) with Kling as a fal
sonic-place-portraits
Produce one 60-second vertical music video for a roots or folk act where a single real place is captured twice in one session: as a 3D Gaussian splat that reacts to the music in the browser, and as the room's own acoustics convolved into the recording. Th
world-labs-3d-generation
Run a structured evaluation of World Labs' Marble (Fei-Fei Li's Large World Model product, closed-source, browser-based) for one concrete job: game-level and previs blockout. Running this produces a prompt set across text, image, and 3D-layout inputs