LLM-powered safety gate for pi: classifies bash and MCP tool calls by risk before execution, with CWD-aware judgments.

Web shell kernel: bootWebShell (module system holding + seed table + two-stage boot + AppRoot gate + app-shell assembly entry), consumed by the apps/web vite entry

A work queue that can be gated, stopping to wait for sub-queues to complete.

Steerable framework — product-neutral renderer SPA source for @steerable/agent-shell (React + Vite). Consumed by product web entry points via the `@/` alias (anchored at this package's src/) and the createProductViteConfig factory. Published as a source p

Autonomous AI agent orchestrator — Claude, GPT, Codex, and local models (Ollama/LMStudio/llama.cpp)

Independent two-layer safety gate around the DeepSeek Harness agent loop: deterministic and model-classifier verdicts for prompts, streamed output, tools, and tool results

Soul-bound workflow roles for Pi

Get accurate and well named css box model information about an Element 📦

Process transparency for agentic work.

Umbrella dispatcher for the @adlc suite - one install, one `adlc <tool>` command surface.

The simplest way to build AI-powered apps

Parser and generator for CSS color strings

Enhanced tool output rendering with syntax highlighting, file icons, tree views, diff rendering, and FFF search

Spawn scoped subagents from within a pi session: agent_run/agent_result tools, Claude Code-compatible agent types, turn caps and tool allowlists

Pi extension for single-agent delegation and scripted multi-agent workflows

Provides React components and integration layer with Adobe Experience Manager Page Editor.

Your agent writes bad React. This catches it

Canonical universal harness — AI-native software engineering via skill-driven orchestration; bootstraps plugkit for task execution and session isolation. Install in any AI coding agent host.

Run an intentic release gate from a CI pipeline, POST what the pipeline knows, wait for the run, exit on the verdict

A cross-harness prose gate for AI agents