my-claude-agents

Bug → spec protocol. When a bug is found or a test fails, trace the cause, decide whether a new §V invariant would catch recurrence, append to §B. This is the one non-obvious thing SDD does that plan-then-execute doesn't. Triggers on test failure, bug report, post-mortem, or explicit user ask.

Plan-then-execute implementation against SPEC.md. Native single-thread loop, no sub-agents. On test or build failure, auto-invokes the backprop skill before retrying — a failed verification always considers whether a new §V invariant would prevent recurrence. Triggers when the user asks to build, implement, execute the spec, or tackle a specific §T task (`build §T.3`, `build --next`, `implement next task`, `run the build`). Expects SPEC.md to exist; if not, defers to the spec skill.

Read-only drift detector. Diffs SPEC.md against current code and reports violations grouped by severity. Writes nothing — suggests remedies via the spec or build skills but never invokes them. Triggers when the user asks to check drift, audit the spec, verify invariants, or ask whether code still matches the spec. Phrasings: "check drift", "audit the spec", "does the code still match §V", "check invariants", "spec vs code".

Create, amend, or backprop bugs into SPEC.md at repo root. Sole mutator of the project spec. Triggers when the user asks to write a spec, start a new spec, distill a spec from existing code, add invariants, amend sections (§G, §C, §I, §V, §T, §B), or record a bug via backprop. Common phrasings: "write the spec for...", "new spec", "bug: ...", "amend §V.3", "distill spec from code", "spec this idea". Reads and follows FORMAT.md for the caveman encoding rules and pipe-table shape of §T and §B.

Decision guide for delegating to caveman-style subagents. Tells the main thread WHEN to spawn `cavecrew-investigator` (locate code), `cavecrew-builder` (1-2 file edit), or `cavecrew-reviewer` (diff review) instead of doing the work inline or using vanilla `Explore`. Subagent output is caveman-compressed so the tool-result injected back into main context is ~60% smaller — main context lasts longer across long sessions. Trigger: "delegate to subagent", "use cavecrew", "spawn investigator/builder/reviewer", "save context", "compressed agent output".

Ultra-compressed commit message generator. Cuts noise from commit messages while preserving intent and reasoning. Conventional Commits format. Subject ≤50 chars, body only when "why" isn't obvious. Use when user says "write a commit", "commit message", "generate commit", "/commit", or invokes /caveman-commit. Auto-triggers when staging changes.

Compress natural language memory files (CLAUDE.md, todos, preferences) into caveman format to save input tokens. Preserves all technical substance, code, URLs, and structure. Compressed version overwrites the original file. Human-readable backup saved as FILE.original.md. Trigger: /caveman-compress FILEPATH or "compress memory file"

Quick-reference card for all caveman modes, skills, and commands. One-shot display, not a persistent mode. Trigger: /caveman-help, "caveman help", "what caveman commands", "how do I use caveman".

Ultra-compressed code review comments. Cuts noise from PR feedback while preserving the actionable signal. Each comment is one line: location, problem, fix. Use when user says "review this PR", "code review", "review the diff", "/review", or invokes /caveman-review. Auto-triggers when reviewing pull requests.

Ultra-compressed communication mode. Cuts token usage ~75% by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.

Show real token usage and estimated savings for the current session. Reads directly from the Claude Code session log — no AI estimation. Triggers on /caveman-stats. Output is injected by the mode-tracker hook; the model itself does not compute the numbers.

Run any question, idea, or decision through a council of 5 AI advisors who independently analyze it from different angles, peer-review each other anonymously, and synthesize a final verdict. Anti-sycophancy design — built to tell you what you don't want to hear. Produces an HTML visual report + markdown transcript. Based on Karpathy's LLM Council methodology. TRIGGER on: 'council this', 'run the council', 'war room this', 'pressure-test this', 'stress-test this', 'debate this'. ALSO TRIGGER on genuine decisions with stakes: 'should I X or Y', 'which option', 'is this the right move', 'I can\'t decide', 'I\'m torn between', 'validate this', 'get multiple perspectives'. DO NOT trigger on: factual lookups, simple yes/no, pure creation tasks (write me a tweet), summaries, or casual should-I without real tradeoffs.

Motion and interaction design expert based on Emil Kowalski, Jakub Krehel, and Jhey Tompkins' techniques. Two modes: build interactive components with purposeful motion, or audit existing animations to catch AI-slop motion patterns (audit emits a branded HTML report with looping demos). TRIGGER when: creating, adding, animating, or reviewing UI motion — transitions, hover states, micro-interactions, enter/exit animations, or any motion design work in React, Framer Motion, CSS, or HTML.

Use when the user wants to design, redesign, shape, critique, audit, polish, clarify, distill, harden, optimize, adapt, animate, colorize, extract, or otherwise improve a frontend interface. Covers websites, landing pages, dashboards, product UI, app shells, components, forms, settings, onboarding, and empty states. Handles UX review, visual hierarchy, information architecture, cognitive load, accessibility, performance, responsive behavior, theming, anti-patterns, typography, fonts, spacing, layout, alignment, color, motion, micro-interactions, UX copy, error states, edge cases, i18n, and reusable design systems or tokens. Also use for bland designs that need to become bolder, loud designs that should become quieter, live browser iteration on UI elements, or ambitious visual effects. Not for backend-only or non-UI tasks. Use /impeccable <command> [target].

Premium brand-kit image generation. Creates high-end brand-guidelines boards, logo systems, identity decks, and visual-world presentations. Trained for minimalist, cinematic, editorial, dark-tech, luxury, cultural, security, gaming, developer-tool, and consumer-app brand systems. TRIGGER when: user asks for brand identity, logo concepts, brand guidelines, identity deck, or visual brand system images.

Anti-slop frontend skill for landing pages, portfolios, and redesigns. Reads the brief, infers the right design direction, and ships interfaces that do not look templated. Real design systems when applicable, audit-first on redesigns, strict 14-point pre-flight check. TRIGGER when: any frontend build, landing page, portfolio, or redesign task. The default frontend skill for all UI work.

The original v1 taste-skill, preserved for projects depending on its exact behavior. The current default is design-taste-frontend (v2), which is a substantial rewrite. Use this v1 install name only if you need exact backward compatibility with projects built on v1.

Overrides LLM truncation defaults. Enforces complete code generation, bans all placeholder patterns, and handles token-limit splits cleanly. TRIGGER when: any task requires full files, complete components, or exhaustive output — especially after partial or truncated responses. Stack-agnostic.

Stricter Awwwards-level frontend variant. Python-driven true layout randomization, AIDA page structure, wide editorial typography (bans 6-line wraps), gapless bento grids, strict GSAP ScrollTriggers (pinning, stacking, scrubbing), inline micro-images, and massive section spacing. TRIGGER when: user wants higher layout variance, stronger GSAP direction, or aggressive anti-slop enforcement beyond the default taste-skill.

$150k agency-level UI engineering. Double-bezel nested architecture, spring physics, magnetic hover, creative variance engine. Polished, calm, expensive UI with softer contrast, whitespace, premium fonts, and spring motion. TRIGGER when: user asks for premium, agency, Awwwards-level, Apple-like, luxury consumer, or high-end brand UI.

Image-first website design pipeline: generate reference images first, analyze them deeply, then implement the frontend to match as closely as possible. Bans lazy under-generation, cards-inside-cards UI, and hero clutter. TRIGGER when: visual quality is the primary concern, user asks for a beautiful/premium/creative website, or any multi-section landing page build where you have image generation available.

Elite mobile app image-generation skill. Premium, app-native screen concepts and flows for iOS, Android, and cross-platform. Prioritizes readable text, multi-screen consistency, controlled color palettes, non-generic creative direction, textured surfaces, image-led composition, and clean phone mockup framing. Generates images only — does not write code. TRIGGER when: user asks for mobile app screen mockups, onboarding flows, app concept images, or mobile UI references.

Elite frontend image-direction skill. Generates one separate horizontal image per section — never compressed into one board. Enforces composition variety, background-image freedom, varied CTAs, hero scale selection, narrative concept spine, and a single consistent palette across all images. TRIGGER when: user asks for website design references, section comps, landing page mockups, or any frontend visual references before coding.

Raw mechanical interfaces fusing Swiss typographic print with military terminal aesthetics. Rigid grids, extreme type scale contrast, utilitarian color, analog degradation effects (halftones, CRT scanlines, bitmap dithering). TRIGGER when: user asks for brutalist, industrial, terminal, Swiss-type, data-heavy dashboard, portfolio, or editorial site that should feel like a declassified blueprint.

Premium utilitarian minimalism — editorial-style interfaces with warm monochrome palettes, typographic contrast, flat bento grids, and muted pastel accents. No gradients, no heavy shadows, no generic defaults. TRIGGER when: user asks for clean, editorial, Notion-style, Linear-style, calm, or document-UI vibes.

Upgrades existing websites and apps to premium quality. Audits current design, identifies generic AI patterns, and applies high-end design standards without breaking functionality. Works with any CSS framework or vanilla CSS. TRIGGER when: user asks to improve, modernize, upgrade, fix the design of, or make premium an existing codebase.

Generates DESIGN.md files optimized for Google Stitch screen generation. Translates anti-slop frontend engineering directives into Stitch's semantic design language — typography, color, component behaviors, layout principles, motion philosophy, and explicit anti-pattern bans. TRIGGER when: user wants to create a DESIGN.md for Stitch, or wants agent-friendly design system documentation for any AI code generator.

Monitor a pull request through CI and review cycles until it is merge-ready. Polls checks, reads review comments, fixes real issues, resolves threads. TRIGGER when: user says "babysit this PR", "watch it until it merges", "monitor the PR", or "keep checking". Do not stop after one pass.

Audit a UI or design against Dieter Rams' ten "Good design is..." principles. Scores each principle with evidence, delivers a verdict (NEW / REFINE / REDESIGN), and hands off a make-plan prompt. TRIGGER when: user says "audit this design", "design review", "critique this UI", "check this against Rams", "is this good design", or wants a design critique with a plan.

Execute a phased implementation plan using subagents. One subagent per phase, one objective per subagent, evidence required before advancing. TRIGGER when: user says "execute this", "run the plan", "do it", or references a make-plan output. Pair with make-plan to plan first, do to ship.

Explains how claude-mem captures observations, when memory injection kicks in, where data lives, and how to verify it's running. TRIGGER when: user asks "how does claude-mem work?", "what is this thing doing?", "is claude-mem running?", "where does it store data?", or "explain the memory system".

Build and query AI-powered knowledge bases from claude-mem observation history. Creates focused "brains" from past session data — all decisions about hooks, all bugfixes in a module, all architecture choices from a month. TRIGGER when: user asks "what did we decide about X", "summarize all work on Y", "build a brain for Z", or wants to query patterns from past sessions. Requires claude-mem to be installed and running.

Prime a codebase by reading every source file in full. Builds deep structural understanding before starting work on unfamiliar repos. TRIGGER when: user says "learn the codebase", "read the codebase", "prime", "get up to speed", "onboard me", or when joining a new project for the first time.

Turn a feature request, task, or multi-session objective into a phased, LLM-friendly implementation plan. Each phase is independently executable by a fresh agent with zero re-explanation. TRIGGER when: user asks to plan, blueprint, or roadmap anything non-trivial. Pair with do to execute. DO NOT TRIGGER when the task fits in a single PR or fewer than 3 tool calls.

Search claude-mem's persistent cross-session memory database for past decisions, solutions, and patterns. TRIGGER when: user asks "did we already solve this?", "how did we do X last time?", "what did we decide about Y?", or needs context from previous sessions. Requires claude-mem installed.

Cluster a GitHub issue backlog by root cause, create plan-master issues for each architectural defect, redirect child issues with standardized comments, and bundle fix PRs that close clusters atomically. TRIGGER when: issue tracker has 20+ open issues with duplicates, user says "triage", "consolidate", "cluster", "dedupe", or "make a roadmap from issues". Also routes new bugs into existing plans.

Map a codebase into feature-grouped flowcharts, identify duplicated concerns across features, and propose the simplest unified architecture. Emits a proposed unified flowchart plus per-system make-plan handoff prompts. TRIGGER when: user says "find the ideal path", "audit architecture", "map the codebase", "unify duplicated systems", or before a major refactor.

Token-optimized structural code search using tree-sitter AST parsing. Replaces the Glob → Grep → Read discovery cycle with smart_search, smart_outline, smart_unfold. TRIGGER when: exploring an unfamiliar module, finding where a function is defined, understanding file structure before editing. Requires smart_search MCP tool to be available.

Generate a "Journey Into [Project]" narrative report analyzing a project's full development history from claude-mem's timeline. TRIGGER when: user asks for a timeline report, project history analysis, development journey, full project report, or "what's the story of this project?". Requires claude-mem installed with session history.

Automated semantic versioning and release workflow for Claude Code plugins and npm packages. Handles version increments across all manifest files, npm publishing, git tagging, GitHub releases, and changelog generation. TRIGGER when: user says "release", "bump version", "publish", "cut a release", or "tag this version".