Unified browser automation for AI agents. Uses surf-cli extension when available (full features), falls back to CDP (zero-config). Navigate, read with element refs, click, type, screenshot.
Universal LLM proxy on localhost:4001. Surfaces: chat/batch completions, scillm exec, OpenCode serve (coding delegate), OpenCode transport (DAG/SSE), standing Codex agents. Chutes, Gemini, Claude/Codex OAuth, OpenCode Go, Ollama. Auto-routes by model name. ZIP/PDF, JSON repair, batch pools.
Zero cognitive-load learning and querying skill. Learn about a topic or persona (e.g., "Lisa Feldman Barrett") by discovering, ingesting, and extracting knowledge — or ask questions against what's been learned. Supports multi-hour deep learning with progress tracking, persona profiles, and nightly incremental updates. Uses Federated Taxonomy for multi-hop graph traversal across knowledge domains. Composes: dogpile, discover-books, ingest-youtube, fetcher, extractor, memory, taxonomy, task-monitor.
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.
Validate ask/scillm DAG JSON and render PHART 1.5 ASCII decision-tree charts for terminals and dry-run output. DAG.json in → chart on stdout or actionable errors on stderr (no tracebacks). Python 3.14+ with PHART from github.com/scottvr/phart.
Best practices for designing and structuring agent skills: SKILL.md frontmatter rules, triggers, progressive disclosure, and when to use scripts vs references.
Force the project-agent to use a real debugger instead of guessing: set breakpoints where the problem might be, stop execution at those breakpoints, inspect live variable state, and analyze the observed runtime state before patching. Use when a project agent is stuck, sees confusing or repeated failures, suspects state mutation, routing, async, serialization, cache, closure, test fixture, UI/backend mismatch, or any bug where logs/static reading would lead to speculation. Use before further patching after two failed attempts or whenever the user asks for debugger, breakpoints, debug mode, variable state, inspect locals, step through, VS Code debugger, or prove runtime behavior.
Evidence-gated phase iteration for implementation plans. Use when a task must proceed phase by phase with deterministic artifacts, validation logs, named external reviewer verdicts, default scillm GPT-5.5 high review, optional reviewer comparison, blocker tracking, and fail-closed acceptance before advancing; especially for security, correctness, deployment, or report-hardening work.