Plugin Registry

Use multiple AI assistants from one interface. Arbiter orchestrates Codex CLI and Gemini CLI inside Claude Code — get parallel answers, take majority votes on decisions, cross-check code, or run 3 independent implementations in parallel and pick the best one.

Publish technical articles to dev.to, Hashnode, Twitter/X, and Hacker News without leaving your editor. Includes a deterministic linter that catches em-dashes, tag format errors, and description length before publishing — plus 10 skills for creating, QA-ing, and distributing content from your existing artifacts.

Autonomous deep research with claim-level verification. Delve runs a 5-stage pipeline (SCAN → DECOMPOSE → DIVE → VERIFY → SYNTHESIZE) that dispatches parallel research subagents, extracts atomic claims, then adversarially verifies each one before producing a synthesis report with quality labels and full source attribution.

Evolve startup ideas through AI-powered natural selection. Three specialized agents run generate-critique-select-refine loops, scoring across 8 dimensions with strategy weights, to surface winners with ICP, monetization model, and week-1 execution plans.

Generate polished terminal demo GIFs from a natural language description. Glyph synthesizes asciicast recordings with realistic typing, configurable themes, and built-in PII auditing — no real shell commands are executed.

Local-first news intelligence for AI agents. Herald collects RSS and Hacker News articles, clusters related stories by title similarity, scores by source weight and recency, and generates ranked Markdown briefs — all locally with zero API keys.

Report bugs, request features, and ask questions for any heurema plugin — without leaving Claude Code. Auto-detects your repo from git remote, collects environment context, previews the issue, and submits via gh CLI with clipboard fallback.

Automated security audit for your AI workstation. Sentinel runs 18 deterministic checks across secrets exposure, MCP trust, plugin permissions, hook coverage, trust boundaries, and configuration hygiene — then scores your security posture 0–100 with guided remediation for every finding.

The code is the easy part. Signum is a contract-first AI development pipeline for Claude Code: spec quality gate (A–F) that blocks weak specs before implementation starts, holdout scenarios the Engineer never sees for blind testing, multi-model audit (Claude + Codex + Gemini in parallel), execution policy enforcement, repo-level invariant contracts, and a self-contained tamper-evident proofpack artifact the AI couldn't game.

Skill telemetry for Claude Code. A PostToolUse hook that logs every skill activation to local JSONL — which skills load, which get followed, how often. The only tool that tracks whether the model actually executed a skill's instructions. Foundation for automated skill quality evaluation and self-improvement pipelines.