Claude Skills by grahama1970

Deck ARCHITECTURE rules for building pitch decks from source material — measured from a real 263-slide corpus, not invented. Use when planning a deck's sections and slide sequence, when deciding which source images become slides, or when reviewing whether a generated deck is house-shaped. Exists to prevent bespoking slide architecture: sections must be DERIVED from the source document, never hardcoded in a compiler.

Best practices for orchestration-ready task files. Enforces adversarial testing, skill chain syntax, definition-of-done requirements, gate definitions, persona routing, and /model directives. Referenced by /plan and /review-plan.

Best practices for porting features from one codebase, app, harness, TUI, API, or workflow into another without losing destination-specific behavior. Use when users ask to port, clone, migrate, replicate parity, achieve functional equivalence, borrow features, compare source vs destination, or preserve custom work while importing behavior from a reference implementation.

Standards and workflow for creating root-level PROJECT_STATE.md reports from local receipts, project knowledge, stale/aspirational audits, competitor research, and external reviewer evidence. Use when asked for project state, comprehensive project review, "what works and what is missing", stale code/doc assessment, competitive positioning, or a PROJECT_STATE.md deliverable.

Project-readiness policy for skills-first repositories and substantial skills. Use when creating, auditing, or declaring a project ready; when deciding which best-practices-* packs apply; when a repo needs setup-project provenance, PROJECT_KNOWLEDGE, explain-project records, cleanup evidence, or retained agentic evals before it can be called developer-ready.

Rules for writing clear, testable LLM prompts. Eliminates vagueness, enforces concrete output specs, grounding in source material, and deterministic verifiability. Gate before /review-prompt.

Repo-specific Python best practices for agentic coding: Loguru + Typer + uv + pyproject.toml, httpx over requests, functions-first, module docstrings, max 800 LOC per file, and non-mocked sanity tests.

React, Next.js, React Native, and web design best practices from Vercel Engineering. Use when writing, reviewing, or refactoring React/Next.js/React Native components, optimizing performance, auditing UI accessibility, or designing component APIs. Triggers: "review react", "optimize component", "check accessibility", "react best practices", "component architecture", "react native performance".

Deploy applications and websites to Vercel. Use this skill when the user requests deployment actions such as "Deploy my app", "Deploy this to production", "Create a preview deployment", "Deploy and give me the link", or "Push this live". No authentication required - returns preview URL and claimable deployment link.

Composition patterns for building flexible, maintainable React components. Avoid boolean prop proliferation by using compound components, lifting state, and composing internals. These patterns make codebases easier for both humans and AI agents to work with as they scale.

React and Next.js performance optimization guidelines from Vercel Engineering. This skill should be used when writing, reviewing, or refactoring React/Next.js code to ensure optimal performance patterns. Triggers on tasks involving React components, Next.js pages, data fetching, bundle optimization, or performance improvements.

Comprehensive best practices for React Native and Expo applications. Contains rules across multiple categories covering performance, animations, UI patterns, and platform-specific optimizations.

Standards and templates for writing concise, welcoming, evidence-aware README.md files for agent-skills skills, public project pages, gallery card destinations, architectural playground repos, and developer-facing docs. Use when creating or reviewing README prose, README structure, project/skill landing pages, image-card destinations, public/private runtime notices, proof and non-claim sections, or documentation meant to be browsed by developers.

Produce evidence-bearing technical reports and report UIs without dashboard theater. Use when creating or reviewing reports, HTML/CSS report surfaces, implementation reviews, design reviews, application inventories, validation readouts, evidence ledgers, status/readiness reports, compliance/control summaries, source-of-truth audits, or issue-to-plan repair reports that must expose persona, primary object, source of truth, findings, rationale, blockers, valid actions, acceptance checks, non-cl...

Canonical rules for writing, formatting, exporting, and verifying Graham Anderson's resume and its public surfaces, so an agent never re-derives them by trial and error. Use when editing RESUME.md, changing the resume PDF, DOCX, Markdown or /resume page, reconciling LinkedIn with the resume, deciding page count or section order, judging whether a claim may be made, or verifying a built resume artifact. /resume executes; this skill decides.

Best practices for leading multi-model Ask roundtables and Tau-DAG deliberation panels. Use when a user asks for a roundtable, model panel, mixed web/API collaborator discussion, multi-round critique, attributed dissent synthesis, or guidance on turning several model perspectives into executable next steps through $ask.

Repo-specific Rust best practices for agentic coding: thiserror + anyhow error handling, tokio async runtime, tracing for observability, clap for CLI, axum for HTTP, serde patterns, cargo workspace conventions, ownership and lifetimes, and idiomatic Rust 2024 edition patterns.

Writing standard for scene scripts and beat descriptions that feed AI video generation: how to describe environment, weather, lighting, and every character/prop interaction with concrete physical states and motion verbs so the persona-dream panel gate passes and rendered clips have life instead of static tableaus. Use when writing or reviewing a scene script, storyboard panel text, dream story beat, or provider prompt whose elements lack physical state, environmental interaction, weather, or ...

Error recovery and anti-patterns for scillm LLM proxy. Load this when scillm calls fail, timeout, or return unexpected results. Covers batch sizing, header requirements, model selection, and debugging workflow.

General application, service, plugin, and agent security hardening baseline. Use before /hack, /security-scan, /code-runner probe generation, or security review. Grounded in OWASP ASVS, OWASP Cheat Sheet Series, Node.js security guidance, and agent/plugin trust-boundary lessons.

Deterministic coded self-improvement loops for *-lab skills. The agent will NOT do this reliably — loops MUST be in code. Proven pattern from Classifier Lab: data_enrichment.py + training_loop.py.

Best practices for designing and structuring agent skills: SKILL.md frontmatter rules, triggers, progressive disclosure, and when to use scripts vs references.

Slide design craft for /pitchdeck: assertion headlines, distance-legible chrome, multi-channel reinforcement, density limits, audience-parameterized arcs, responsive browser presentation, and house theme templates. Use when planning, drafting, critiquing, or restyling a deck; when slides look like "blue boxes"; when choosing layouts, fonts, or themes; or when converting a reference PPTX into a house style.

Best practices for the SPARTA pipeline: data model, edge persistence, QRA generation, ArangoDB collections, and common gotchas learned the hard way.

Project-specific best practices for the Stream Deck control suite: icon format constraints, socket vs config boundaries, widget button lifecycle, page creation via build_page, ArangoDB fallback patterns, and NVIS palette compliance.

Canonical contract, tool-policy vocabulary, memory/ToM-lite rules, helper delegation, receipt requirements, and retry budgets for Scillm/OpenCode/$loop subagents and persona workers.

Best practices for authoring, reviewing, and repairing Tau DAG contracts. Use when creating tau.dag_contract.v1 YAML/JSON, choosing project-agent subagent roles, declaring immutable goals, adding skill gates such as best-practices-prompt/react/python, specifying provider/model policy, or diagnosing project-agent DAG failures and tau.dag_error.v1 course-corrections.

Guided onboarding walkthrough for new Embry OS users via Embry Lawson persona

Free web, local, LLM Context, and optional Summarizer search via Brave Search API. Use when user says "brave search", "search with brave", "brave web search", "brave local search", "local search", "find businesses near", "near me", "Brave LLM context", or "Brave summarizer".

Persistent browser-oracle tab bindings and directory walk-up registry for WebGPT, Cursor Browser, Gemini, Kimi, and Claude. Binds tab id / viewId + URL once under ~/.pi; maps directories to project names via .ask/browser-oracles.yaml discovered like python-dotenv parent walk-up.

Authorization-gated CAPTCHA security evaluation using the pinned ReCAP agent, synthetic dynamic CAPTCHA challenges on loopback, Surf transport proof, and Ask DAG composition. Use when the user says "evaluate CAPTCHA security", "benchmark ReCAP", "test a local CAPTCHA agent", or "measure CAPTCHA robustness".

Research and decide visual casting for persona-dream/movie entities before contact-sheet generation. Use when a story, screenplay, storyboard, or visual_entities.json needs characters, creatures, props, environments, moods, time periods, states, or user-provided reference images turned into a casting_contract.json for downstream contact sheets. Use for "casting agent", "visual casting", "choose references", "research character look", "use these image paths as references", "Brave-search for re...

Copy the last complete Cursor user/assistant turn to the clipboard (Codex-style /copy). On modern Cursor Agent installs reads ~/.cursor/projects/*/agent-transcripts/*.jsonl when SQLite bubbleId rows are absent. Use for ccopy, cursor copy, or export last Cursor turn.

Thin front door for speaking a line through the live Chatterbox service with a named voice, tone, context note, and emotional intensity. Use when a project agent says "speak", "say this out loud", "speak as embry", "chatterbox speak", "render this with a sigh", or wants one-shot voiced output with a receipt without the full embry-voice-control turn pipeline.

Session grading and cataloging for the learning flywheel. Stores checkpoints in ArangoDB via httpx to memory daemon — findable by BM25, semantic search, and multi-hop graph traversal. Grades sessions with a 5-level rubric, stores proven/failed skill chains for /recommend-skill-chain, links to episodic archives, and cross-references ~/.claude/ memory. Git commits both project AND skills on every save. NON-NEGOTIABLE.

Centralized Chutes.ai LLM gateway as a shared Docker service. Global 5-slot semaphore enforces account concurrency limit across all agents. Single and batch calls with tenacious retry mode, circuit breaker, fallback chain, NDJSON streaming, and structured request/response objects with JSON cleaning.

Validate subagent reliability by running 3 concurrent classifier training loops. Each subagent runs /classifier-lab (deterministic code) and reports via JSON stream. The subagent's only job: run the skill, fix crashes, write /test-lab tests. 90% code, 10% agent. HP adjustment between rounds uses /scillm Gemini structured JSON, not agent judgment.

Multi-modality classifier training lab. Train, evaluate, and deploy vision, text, and tabular classifiers using state-of-the-art backbones.

Normalize text to handle PDF/Unicode encoding issues and security evasion. Converts Windows-1252, curly quotes, em/en dashes, ligatures, directional formatting, zero-width chars, and more to clean ASCII. Security extensions: homoglyph translation, mixed-script detection, invisible content detection, URL double-decoding, letter-spacing collapse.

Assess the project to reorganize or deprecate unused/outdated files. Joins each candidate against per-candidate dependency evidence, blocks mutation per class on the evidence that class actually needs, and keeps assessment running when indexing is unavailable.

Copy files, zips, images, screenshots, artifacts, or bundles to the desktop clipboard as file-manager paste items. Use when the user says copy a file to clipboard, copy a zip to clipboard, copy an image to clipboard, put this artifact on the clipboard, or when KDE/X11, GNOME, xclip, text/uri-list, or x-special/gnome-copied-files clipboard targets matter.

Copy files, bundles, zips, images, screenshots, artifact paths, or text to the desktop clipboard with verified KDE/GNOME-compatible MIME targets. Use when the user says "$clipboard", "copy to clipboard", "put this on clipboard", or asks to copy a file/artifact/zip/image/bundle.

Assess CMMC Level 2/3 compliance by mapping NIST SP 800-171 controls to Embry OS features and detected configurations. Generates gap analysis reports.

Deterministic code review skill with T0 validators (best-practices-*, ruff, compile) and LLM-powered findings (codex/scillm). Scores findings by severity, keeps suggested fixes advisory, and fails closed on provider errors. Structured JSON output. Replaces raw codex exec in orchestrate T2 gate.

Minimal deterministic code runner for bounded implementation tasks. Runs one LLM backend through /scillm inside a disposable git worktree, enforces a write allowlist, runs a deterministic definition of done, and returns an allowlist-scoped patch artifact for project-agent review.

High-reasoning agentic bridge via OpenAI Codex CLI. Supports gpt-5.3-codex with optional reasoning effort. Use for complex analysis, code generation, and structured extraction. Watchdog stall detection and task-monitor integration for long-running calls.

Chronological audit timeline of compliance changes from ArangoDB revision history

Search and annotate ingested books. Supports markdown/text files and EPUBs, tracks reading position, and stores Horus notes in memory.

Manage and run nightly ingestion of upstream feeds (RSS). GitHub and NVD support coming in Phase 2. Fetch updates, store summaries in ArangoDB, and integrate with Memory. Use this skill to "check for updates" or "add a new source" to the knowledge graph.

Search and retrieve MIDI fragments from the library for arrangement composition. The library is populated from reference songs: /consume-music finds songs → /learn-artist downloads + separates stems → /review-music analyzes → /create-midi to-spec converts stem audio to MIDI → /memory learn stores with HMT tags. Queries by key, BPM, instrument, mood (heart tags).