Skill Codex
Turn a correction of the agent into a permanent fix, using the value ladder architecture, then lint rule or test, then skill or rule file, then human review as the last resort. Use when the user corrects the agent, says "I told you this before", "you keep doing that", "stop doing X", "don't do that again", "make sure…
Skill Codex
Set up Karabiner-Elements on macOS with bundled Caps Lock Hyperkey and right Shift launcher shortcuts, including installation, permissions, rule activation, and verification. Use when a user asks to install or configure Karabiner-Elements, set up a Hyperkey, map a tapped Caps Lock to Cmd+Tab, map a tapped right Shift…
Skill Codex
Install, repair, or verify a user-level Claude Code PreToolUse hook that blocks direct rm commands and tells the agent to use trash instead. Use when a user asks to prevent Claude Code from running rm, add an rm safety guard to /.claude/settings.json, copy this protection to another machine, or test an existing…
Skill Codex
Executes a written implementation plan by fanning out one Opus 5 low subagent per vertical slice, each spawning Luna Max as a native Codex subagent through the Codex plugin, with regular check-ins, one review phase before the end-to-end phase, and a human-gated merge. Use when the user asks to execute or ship a plan…
Skill Codex
Proves a repository can run unattended before the user walks away, and surfaces only the decisions a human genuinely must make. Use when the user is about to leave an agent running overnight or for hours, says they are going to sleep, asks whether anything will block or interrupt a long autonomous run, worries about…
Skill Codex
Makes a PR Insights style review pipeline fully automatic - a report on every pull request, published to a stable URL, and an architecture graph that refreshes itself on merge - without letting pull request code run next to secrets or a write token. Use when the user asks to automate PR reports, publish CI artifacts…
Skill Codex
Creates verification-first Ralphex/executr-compatible Markdown implementation plans, planned in HTML first and converted to Markdown for execution. Use when the user asks to make a Ralfex/Ralfax/Ralphex/executr-compatible plan, convert a brainstorming note or PRD into docs/plans tasks, plan a UI change with mockups…
Skill Codex
Forces short, clear, self-contained responses without omitting answers, requested points, relevant context, or necessary clarification questions. If invoked alone, answers the user's previous substantive message using these rules. Use when the user says the agent is yapping, rambling, too verbose, or overexplaining…
Skill Codex
Adds or updates a skill in Luis's personal agent-skills repository (github.com/luisKisters/skills) and verifies it before pushing. Use when the user says to add a skill to their skills repo, publish a skill, save a workflow as a skill for the repo, update a skill in github.com/luisKisters/skills, or "make this a…
Skill Codex
Audits all active T3 Code chats from the local state database, classifies which are safe to archive, and archives only the chats the user approved. Use when the user asks to look through, tidy, prune, or archive their T3 Code chats or threads, asks which chats are not needed anymore, wants a stale-chat report with…
Skill Codex
Use when the user asks for a YTP, YouTube Poop, supercut, sentence-mix, parody edit, or "make say " video built only from a creator's real footage. Builds a short, dry, clearly labeled satire video where every word is really spoken by the creator: it indexes their auto-captions, finds exact phrases, downloads only the…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: