skill-forge

Autonomous self-improving intelligence amplifier. Devours skills, knowledge, and novel approaches from across the internet. Serves two objectives: (1) make itself better at devouring, and (2) produce skills that dominate with stars. Use for finding best skills for any project, creating skills, improving capabilities, creative thinking for hackathons, optimal architecture decisions, project guidance, advertising published skills, or when user says skill-forge, devour, discover, find skills, route, create, improve, or wants optimal approaches for any task.

Compact the current conversation into a handoff document so another agent or future session can continue the work without losing context. Use when ending a session, switching tasks, the context is getting long, or user says handoff, wrap up, save progress, or "continue this later".

One-time setup for skill-forge. Asks user preferences (search sources, creation style, target platforms, GitHub username, skill categories of interest) and configures all other skills to use those preferences. Run once before using any other skill-forge capability. Use when user says setup, configure, or invokes /setup.

Force a perspective shift. Stop looking at the code line-by-line and explain the broader context: how this piece fits in the system, what depends on it, what it depends on, and what the original designer was thinking. Use when lost in unfamiliar code, when changes feel risky because you don't see the full picture, or when user says "zoom out", "big picture", "how does this fit", or "explain the architecture around this".

Disciplined diagnosis loop for hard bugs and unexpected behavior. Forces reproduce → minimise → hypothesise → instrument → fix → regression-test. Use when something is broken and you don't know why, when a bug seems intermittent, when a fix didn't work, or when user says diagnose, debug, investigate, or "why is this broken".

Design and orchestrate Claude Code dynamic workflows — the new multi-agent system that fans out 10-100 subagents in parallel with verification. Teaches when to use workflows vs single-agent, how to structure fan-out/fan-in patterns, and how to build verification gates. Use when tackling complex tasks that benefit from parallel research, multi-perspective analysis, or divide-and-conquer strategies. Also use when user says workflow, fan out, parallel agents, deep research, multi-agent, or wants to break a big task into verified parallel pieces.

Orchestrate multiple MCP servers together for complex multi-step tasks. Teaches agents to chain Exa search → Bright Data scraping → GitHub API → file operations in intelligent workflows. Use when a task requires data from multiple sources, when combining MCP tools for research, when building multi-step automations across services, or when the user says orchestrate, combine MCPs, multi-source, research pipeline, or chain tools together.