3,748 tools found
codex-profiles gives Codex CLI and ChatGPT desktop separate named local homes per account, without copying or inspecting tokens.
Runs a read-only local security audit for a CVE or advisory and writes a dated, permanent Markdown report.
Distributes an authored skill to all 4 local agent skill folders - Codex, Claude Code, Pi, and Hermes - keeping them in sync.
Defines the Agent Skills open standard - SKILL.md structure and the three-level progressive-disclosure loading model across agents.
Reads a folder in full, drafts an outline for approval, then writes a folder-scoped CLAUDE.md plus a synced AGENTS.md symlink.
Multi-agent research harness: parallel web searches, source fetching, 3-vote adversarial claim checks and a cited report.
Turns an agent prompt into a persistent /goal loop that plans, acts, tests, and iterates until a verifiable stop condition.
Delegate scoped implementation tasks to Grok Build with explicit disclosure approval, clean worktrees, diff review, and acceptance checks.
Builds durable strategic docs one question at a time - AI asks, the user answers, AI patches, never proposes content.
Registers an unrecognized OpenRouter model slug (like a :nitro variant) as a custom Pi model so it stops silently falling back.
Pilot Protocol gives AI agents a permanent network address, encrypted P2P tunnels, and an app store of typed local capabilities.
Executes safe, read-only PostgreSQL queries across multiple configured connections with layered write protection.
