toolu

Dual-host engineering discipline for Claude Code and Codex: workflow skills, agents, commands, and registry-driven quality hooks.

7 skills · 5 commands · 6 agents · hooks

Drive a GitHub, Jira, or Linear epic to merged PRs: dependency graph, herdr worktrees, parallel Claude Code / Codex / Cursor Agent workers routed by Jev complexity, merge gate, rate-limit and progress guardrails.

1 skill · 1 command · hooks

Babysit a PR until review threads, the review-bot verdict, and CI are clear, using Claude cron or a durable Codex goal; fixes run on Jev-routed Claude Code, Codex or Cursor Agent sessions in a herdr worktree.

1 skill · 1 command · hooks

Native Claude Code statusline and Codex status skills for repository and gate status. Standalone, no dependencies.

1 skill · 1 command · hooks

Structural code search and rewrite: skill, native toolu ast-grep commands, and compiled search and byte-savings rules.

1 skill · hooks

Native code review of branch and uncommitted changes with host CLIs and Jev. Produces anchored findings, a compact verdict, review memory, and gate-compatible push state.

1 skill · 1 command

Keep documentation accurate with a Jev-routed editing sub-agent and native review records.

1 skill · hooks

Create and refresh GitHub epics with cited decisions, coherent sub-issues, observable acceptance criteria and test plans. Includes epic-creator and epic-updater, with graph verification before worker handoff.

2 skills

Semantic branches and commits agents can read: scoped Conventional Commit subjects with an Intent trailer and <type>/<ref>-<slug> branch names, checked before git runs.

1 skill · hooks

Typed judgments from TypeSafe's Jev model at runtime (skill + REST wrapper) — yes/no probabilities, choices, and scores over text and images that a script can branch on. Standalone, no dependencies.

1 skill · hooks

Evidence-backed review of committed changes for actionable defects, with complete-coverage push-review state. Standalone native-review fallback.

1 skill · hooks

Think a change through before building: evidence-backed triage, alternatives, trade-offs, and a recommended approach, without editing code. Standalone, and phase 1 of delivery-flow. No dependencies.

1 skill

One dual-host delivery skill for brainstorm, approved spec and plan, real-data execution, PR creation, and pr-babysit handoff.

1 skill

Python PostToolUse quality checks registered into the toolu hook engine.

hooks

Rust PostToolUse quality checks registered into the toolu hook engine.

hooks

A Linux sandbox for builds, gates and agent CLIs: a Lima VM on macOS that mounts only the project, this machine on Linux. `toolu sandbox setup` provisions the toolchain, and `toolu sandbox run` runs any command there at the same path. Standalone, no dependencies.

1 skill

TypeScript PostToolUse quality checks registered into the toolu hook engine.

hooks