Loading...
Loading...
Found 123 Skills
Self-improving browser automation via the auto-research loop. Iteratively runs a browsing task, reads the trace, and improves the navigation skill (strategy.md) until it reliably passes. Supports parallel runs across multiple tasks using sub-agents. Use when you want to build or improve browser automation skills for specific website tasks.
Propose and execute rubric or bucket upgrades. Two modes: **Full rubric bump** (highest-risk action, mandatory 5-step process + cross-model audit) and **--bucket-only lightweight recalibration** (only update bucket boundaries, no changes to rubric formulas). **Phase 2 mandates using cheat-score-blind sub-agent to re-score the calibration pool** — self-scored fallback is not accepted. Trigger phrases: "upgrade rubric"/"bump rubric"/"update formula"/"I want to add a dimension"/"adjust weights"/"recalibrate bucket"/"recalibrate bucket".
Canonical Claude Code authoring kit covering Skills, sub-agents, plugins, slash commands, hooks, memory, settings, sandboxing, headless mode, and advanced agent patterns. Use when creating Claude Code extensions or configuring Claude Code features.
Stay current with how OpenCode, OpenAI Codex, and Claude Code implement extensibility features (skills, slash commands, subagents, custom prompts). Use when comparing implementations across AI coding assistants, researching how a specific tool implements a feature, or syncing knowledge about agent extensibility patterns. Triggers include questions like "how does X implement skills?", "compare slash commands across tools", "what's the latest on Claude Code sub-agents?", or requests to understand agent extensibility approaches.
Semantic search, context management, and document indexing via OpenViking. Use when the user asks to: index/import documents or files into a knowledge base, perform semantic search across indexed content, browse or explore indexed resources, get summaries/overviews of indexed documents, manage an OpenViking instance, or integrate structured context retrieval into workflows. Also use when sub-agents need to retrieve relevant context from a large document collection.
Plan and execute large refactor or rewrite efforts efficiently with parallel multi-agent analysis and implementation. Use when a user asks to refactor many files, split workstreams, analyze a target code area, and coordinate sub-agents with clear ownership and dependency-aware execution.
Use this skill for ANY multi-pane or multi-agent terminal orchestration in cmux. Required when the user wants to: run things in parallel in separate terminal panes, split the terminal, spawn a sub-agent (Claude Code, Codex) in another pane, fan out tasks across splits, send keystrokes or text to another pane (including ctrl-c), read terminal output from another pane, update sidebar status or progress bar, open a URL in cmux's built-in browser pane, or display markdown preview alongside the terminal. The cmux CLI is the ONLY way to do these things — Bash cannot split panes or spawn agents. Trigger phrases: 'in parallel', 'split pane', 'spawn agent', 'fan out', 'new pane', 'browser pane', 'sidebar', 'send to pane', 'read from pane', 'show the plan', 'ctrl-c to', '分屏', '并行', '开个 pane'. NOT for: single command execution, basic bash operations, or questions about tmux.
Explore a codebase for architectural friction, discover refactoring opportunities, and propose module-deepening refactors as GitHub issue RFCs. Uses friction-driven exploration and parallel sub-agents to design multiple interface alternatives. Use when user wants to improve architecture, find refactoring opportunities, consolidate coupled modules, reduce complexity, make code more testable, or review codebase health.
This skill should be used when the user asks to "design multi-agent system", "implement supervisor pattern", "create swarm architecture", "coordinate multiple agents", or mentions multi-agent patterns, context isolation, agent handoffs, sub-agents, or parallel agent execution. Part of the context engineering skill suite — also activates when the user mentions "context engineering" or "context-engineering" in the context of orchestrating context across multiple agents.
Use when the user asks for a broad codebase review, substantial PR/branch review, architecture audit, tech-debt scan, cleanup assessment, structural sanity check, or design-alignment review. Default workflow: use sub-agents when available unless specifically forbidden; do not require the user to mention sub-agents, council mode, delegation, or parallel review. Focus on cruft, duplication, weak boundaries, missed reuse, lifecycle/concurrency risks, test/roadmap drift, and code aesthetics. Do not use for narrow bug fixes, ordinary small-diff reviews, frontend visual QA, repo-onboarding docs, or OpenAI Agents SDK production-readiness review. Output evidence-backed findings first, then pressure points, design alignment, open questions, and follow-through.
Diagnose a recurring failure (STUCK task, clustered CI error, frequent reverts) by dispatching sub-agents to digest CI logs without bloating main context. Returns one root-cause diagnosis.
Adopt multiple expert personas sequentially for complex problem analysis from diverse perspectives. Single-agent only — do NOT spawn sub-agents.