AI agent

Agent Skill that hunts hallucinated packages, APIs, files, and fake tests before a coding agent can claim done. Claude / Cursor / Codex / Copilot / Gemini. npx skills add Jaysi88/ghost-check

The intelligence layer for beads: typed, calibrated labels for your backlog

Throwaway Jev demo: route coding tasks to Grok Build or Codex Astra

jevil is in the details

Benchmark of 6 coding-agent harnesses (omp/pi/fx/opencode/dsh/crush) as headless agent loops driven by a control plane. Two tiers: static source audit + live runs.

An OpenCode V2 skill that lets an agent in one session send a request to an agent in another session — and get an explicit reply back. Markdown only, no plugin. Verified against a running 2.0.6 server.

An open-source AI judge for claims, plans, code, and decisions

Self-hosted, versioned skills library for AI agents. MCP, scoped clients, and optional Jev recommendations.

Some personal thoughts, turned into blog posts (short or long) after being processed by an agent. These aren’t rigorously reviewed blogs; the main goal is to express ideas, so the wording is a bit casual. If there are any errors, just treat them as AI hallucinations. Just posting casually, keeping them as credit for witnessing the arrival of AGI.

a lean mean worker in lean mean lean

Catch AI coding-agent scope creep before you commit. Overscope compares Claude Code and Codex sessions with your git diff to flag out-of-scope edits, deleted tests, hidden dependencies, and false completion claims. Local, deterministic, no model or API key.

Bounded TypeSafe Jev workflows for coding agents.

Model-agnostic project memory for AI agents: resume from verified task state, not chat history.

Agente experimental para explorar hasta dónde un LLM puede cubrir tareas de conversación psicológica: detectar patrones, formular hipótesis, cuestionar interpretaciones y razonar con incertidumbre. Proyecto académico; no sustituye evaluación profesional.

A cyberpunk border encounter powered by TypeSafe Jev. Bluff the guard. Inspect the receipts.

A collection of 7 custom Claude Skills to improve agentic workflows, reasoning, and output quality.

Evaluating TypeSafe's Jev as a fast monitor and action gate for agent sabotage in SHADE-Arena, compared with Gemini 2.5 Flash/Pro.