ellyseum

Interactive setup wizard for claude-vision - configures Docker images and platform settings

Read clipboard contents (text or image) and respond to a prompt about it

Visual research - search the web, find images, analyze them visually

Read the latest screenshot and respond to a prompt about it

Analyze a video or screen recording by extracting frames and spawning an analysis agent

Load ALL memory files into context (nuclear option for planning sessions)

Display the memory file knowledge map tree

Force rebuild the knowledge map cache

Autonomous looped repo work. For each work unit (phase, bucket, item, whatever the user calls them): plan → peer-review council → revise (≤2 rounds) → implement → test → diff council → commit. Background-ping wake-up between iterations. Repeat until the user's roadmap is exhausted or a halt trigger fires. PERSISTENCE: Claude Code's task tool is the primary state — work units and per-unit grinder steps are TaskCreate'd at run start, TaskUpdate'd on every transition. Tasks survive auto-compact, so wake-up reads the task list to find where it left off. ACTIVATE on: "grind the backlog", "grind through this", "auto-grind <X>", "run the grinder on <X>", "go autonomous on <X>", "continue with the <buckets|phases|items>", "use the auto repo grinder". REQUIRES: `cordy` on PATH and at least 2 peer drivers reachable. Refuses to start with fewer drivers — single-peer "council" defeats the diversity benefit. Auto mode (or equivalent permission-bypass) must be active; the loop cannot pause for permission prompts. NOT f

Drive `cordy` (the @ellyseum/cordyceps CLI) to consult peer LLMs from different model families — Codex (OpenAI), Gemini (Google), Ollama (local), and additional Claude variants — for ANY task where cross-model input improves the answer: code reviews, plan critiques, red-teaming, brainstorming, decision exploration, claim cross-checking, meta-reviews of reviews. Capability skill — Claude picks the right shape (single peer query, silo'd panel + chair, adversarial pair, divergent panel, tiered escalation, persistent sidecar, meta-review) for the user's ask. ACTIVATE on explicit cross-model phrases: "consult peers", "spawn council", "convene council", "ask the panel", "code review with peers", "plan critique with peers", "brainstorm with peers" / "brainstorm with multiple LLMs", "what would <name-of-model> think about this", "ask Codex/Gemini/Ollama". TIGHT, NOT BROAD: phrases like "second opinion", "peer review", "plan review", "red team this" are AMBIGUOUS — they can mean "Claude take another pass" rather tha

Securely execute commands requiring API keys without exposing secrets in session logs

Externalize Claude's auto-memory into your project directory so it's git-committable and survives across machines