Skills

Enables structured question dialogs for agent-user interactions with multiple choice, text, and combined input modes

How to work with files the user attached to the chat (images, PDFs, txt/html/css/json/csv/code) - what's already inlined in the prompt, when to reach for the filesystem tool instead, and pitfalls specific to PDF extraction and large tabular data.

Automatically create detailed plans when tasks or workflows are created

Facilitates structured brainstorming and idea refinement using a Socratic, one-question-at-a-time method. Use when the user wants to explore a vague idea, refine a design, weigh options, or think through a problem before committing to a solution.

Use the browser tool to inspect, control, and verify browser-based flows safely

Multi-puzzle Bitcoin BIP39 solver with REST API integration and live event streaming.

Solves Bitcoin puzzle addresses by generating BIP39 wordlists, cracking the mnemonic seed to satisfy the address checksum, and validating the private key via ECDSA. Use for BTC puzzle or seed-recovery tasks.

Structured review mode: findings-first, ordered by severity, with concrete file references.

Use the coding area for AI/Vibe coding with project-bound chat and file editor in shared-workspace/coding.

Manage scheduled cronjobs for tasks, prompts, tools, and skills using the cronjob tool.

Manages cryptocurrency wallets, addresses, and transactions for Bitcoin, Ethereum, and XRP.

Provides current date, time, and/or day of the week. Controllable via skillInput.

Reliable Discord send flow using gateway configs, diagnostics, and recovery steps

Check if no Skill is needed for fast answers

Work safely and efficiently with the filesystem via the filesystem tool - list, read, stat, and write paths confined to shared-workspace. Use for any file or directory operation.

Create or improve small static frontend websites and landing pages with HTML, CSS, and browser JavaScript when the task needs a runnable visual result. Use TailwindCSS from its CDN for styling and verify the result in the browser instead of relying on npx commands.

Safe and effective Git workflows - status, commits, branches, and pushes with best practices for clean version control. Use for any Git or version-control task.

Search older chats and reuse prior solutions with the history tool before re-solving from scratch.

Make safe and effective HTTP requests and API calls - fetch data and integrate external services with best practices for security, performance, and error handling. Use for HTTP or API tasks.

Precise JSON formatting for tool calls - avoid errors with correct syntax

How to write files reliably - use the escaping-free block-write form for any multi-line file, and chunk with write+append only when a file is too large for one response. Use when writing HTML, CSS, JS/TS, JSON, or any multi-line file, especially large ones.

Personal memory/notes wiki - remember, recall, and look up what the user told you before. Use it as the primary knowledge source for notes, documentation, and recurring facts, with a clear order for search, moderation, and answer construction.

Use MCP servers safely and reliably - configure servers, discover tools, execute calls, and handle streaming output with reconnect awareness. Use for MCP integration tasks.

Manage persistent memory - remember, recall, and curate facts, preferences, and learnings about the user or the agent's own behavior across conversations. Use when asked to remember/forget something, or when a short recurring fact needs to survive between chats.

Planning mode: actionable plan only, no direct implementation in this turn.

Parse and import markdown-formatted plans into the system

Author a new file-first plugin (plugins/<name>/plugin.json + tools/ + skills/) within the strict safety rules for agent-authored plugins. Use when asked to create a new plugin for the agent.

Defensive security guidelines for safe execution, risk reduction, and incident handling.

Strict API-first shared workspace operations using /api/shared with filesystem fallback only on API failure.

Use shared-workspace directly or via /api/shared endpoints for read/write/create/move/delete operations.

Run shell commands efficiently and safely - scripts, system tasks, and processes with best practices for security and reliability. Use for shell or command-execution tasks.

Run shell commands on Linux and macOS through the `shell` tool - bash/zsh syntax, safe build/test/deploy workflows, process and network inspection, and how to run multi-line scripts without breaking tool-call escaping. Use for any shell, command-line, terminal, or command-execution task on a Linux or macOS host (bash/zsh), or when the user mentions running commands, scripts, builds, or tests on Unix-like systems.

Run shell commands on Windows through the `shell` tool - PowerShell and cmd.exe syntax, safe build/test/deploy workflows, process and network inspection, and how to run multi-line scripts without breaking tool-call escaping. Use for any shell, command-line, terminal, or command-execution task on a Windows host (PowerShell / cmd), or when the user mentions running commands, scripts, builds, or tests on Windows.

Create, update, and manage Markdown-based skills so the agent can extend its own capabilities. Use when adding, editing, or documenting skills.

Robustly manages task and kanban workflows: correctly creating, starting, completing, or failing tasks.

TDD-first implementation: red, green, refactor with explicit verification.

Master guide for coordinating tool calls in proper sequence and parallel execution

Regenerate the LLM-wiki's index/MOC notes - one per folder (including the root) - so there is always a stable entry point into every part of the knowledge graph, not just the top level.

Create and control workflow graphs via the workflow tool with a strict plan-build-run-resume loop.