Skill Claude CodeCodex
Run and manage long-lived project processes (rails/puma, npm, mix phx.server, cargo run, docker compose, workers) in tu virtual terminals so the human and the agent watch the same screen. Use when: start the server, run the app locally, boot the stack, restart the backend, is it still running, why did it crash, show…
Skill Claude CodeCodex
QA engineer that smoke-tests changes in the browser, validates behavior, plans fixes or test gaps, and updates documentation. Acts as a "second dev" focused on testing what was built. Use when ready to validate work before committing. Trigger on phrases like "qa", "smoke test", "test this", "validate", "check my…
Skill Claude CodeCodex
Save session learnings to the Obsidian vault. Use at the end of a work session to capture what was learned, decided, or discovered. Trigger on phrases like "recap", "save session", "what did we learn", "wrap up".
Skill Claude CodeCodex
Code review in 2 modes. Autonomous (default): spawns parallel security/performance/quality reviewers + red-team auditor, presents verdict. Pair: walks changed files one-by-one with you, collects comments, posts only on approval. Accepts prompt, GitHub/Linear URL, PRD path, or no args. Use when: review, code review…
Skill Claude CodeCodex
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
Agent Claude Code
Analyze blind comparison results to understand WHY the winner won and generate improvement suggestions.
Agent Claude Code
Compare two outputs WITHOUT knowing which skill produced them.
Agent Claude Code
Evaluate expectations against an execution transcript and outputs.
Skill Claude CodeCodex
Discovery before building. Reproduce and observe real behavior, collect evidence, close a mini bag of 1-3 small issues. Use when the ask is fuzzy: a vague bug, "is this at parity?", a feature idea, "migrate X". Feeds /step or /pair, which ride the bag.
Skill Claude CodeCodex
Work in the smallest next step, staying in the loop. Browser-first RED, TDD, evidence, before/then. Handles refining a requirement, researching alternatives, or building — and switches between them as the task shifts. Keeps a tiny optional buffer. Output stays short and critical. Use when: step, let's build this, next…
Skill Claude CodeCodex
Scout the current project, diff against existing CLAUDE.md and .claude/rules, propose updates to reflect stack and conventions. Idempotent. Does not copy user-level skills or agents (user-level always wins). Use when: sync, forge, tune, init project, bootstrap, prepare project, refresh rules.
Skill Claude CodeCodex
Test TUI applications using tu (terminal-use). Launch apps in virtual terminals, take screenshots, type text, press keys, wait for conditions, and verify output. Use for smoke testing terminal UIs, interactive CLIs, and TUI apps. Trigger on phrases like "test the TUI", "terminal test", "tu test", "test in terminal"…
Skill Claude CodeCodex
Search and retrieve notes from the Obsidian vault. Use when the user wants to find notes, recall past ideas, get context on a topic, look up references, browse blog drafts, get live stream ideas, consolidate notes for writing, or load knowledge from their second brain into the conversation. Trigger on phrases like…
Skill Claude CodeCodex
Navigate a code review in chunks, with the user. Works on a PR (theirs or a colleague's) or on uncommitted work, across repos. Each chunk is one reviewable decision: surface affected, before/after, the call made, security/perf/quality flags, and blind spots. Feedback accumulates in an annotation bag; ends in a staged…