kronael

tools

Command-line utilities and Claude Code configuration.

CLI tools

• udfix — fix Unicode box-drawing junction chars in ASCII diagrams (stdin → stdout)
• dockbox — dockerized Claude Code sandbox
• rig — ripgit: smart branch checkout, push, rebase, merge
• tw-fetch — Twitter/X thread archiver
• tg-fetch — Telegram channel/group archiver (telethon, TOML config)
• dc-fetch — Discord channel archiver (discum, DISCORD_TOKEN env)
• clp — claude project picker (experimental; sourceable bash function)

Go tools (udfix, rig): cd <tool> && make install. PEP 723 scripts (tg-fetch, dc-fetch): uv run main.py. dockbox has its own Makefile.

External tools used by the Claude Code config:

• ship — planner-worker-judge CLI
• agent-browser — headless browser automation

Claude Code toolkit

Claude plugin path:

/plugin marketplace add kronael/tools
/plugin install kronael@kronael
/kronael:install

This runs kronael/install/SKILL.md, the single source of truth for the procedure. The install step exists so your ~/.claude/ becomes a working copy you can edit and PR back; pure plugin updates would overwrite edits. Full rationale: ARCHITECTURE.md.

Codex installer bridge

Codex does not run Claude Code slash commands, but it can install this same Claude bundle. The Codex plugin exposes one skill, kronael-install, whose job is to read kronael/install/SKILL.md and run the install path from the GitHub marketplace snapshot.

The Codex plugin is only the installer bridge. It contains one Codex skill (kronael-install); the Kronael bundle installs to ~/.claude/, then the bridge exposes those installed skills to Codex through ~/.agents/skills.

Codex plugin path:

codex plugin marketplace add kronael/tools
codex plugin add kronael@kronael

Then start a fresh Codex thread and ask:

Use $kronael-install to install/update Kronael.

The same skill also handles bridge-only setup:

Use $kronael-install to bridge CLAUDE.md and .claude/skills into Codex.

The bridge does not duplicate skills/, agents/, or hooks/ into the Codex plugin cache. It reads the GitHub marketplace snapshot, deploys the Claude bundle into ~/.claude/, and symlinks the installed skills into Codex's user skill location.

To repair or apply only the Codex side of that bridge, ask:

Use $kronael-install to bridge .claude/skills into Codex.

That links ~/.agents/skills to ~/.claude/skills when possible. If ~/.agents/skills already exists as a directory, the bridge adds per-skill symlinks for Kronael skills instead. Codex scans ~/.agents/skills, not ~/.claude/skills.

Codex compatibility for Claude projects:

• Add CLAUDE.md to project_doc_fallback_filenames in ~/.codex/config.toml so Codex reads Claude-only project instructions.
• If a project already has AGENTS.md, keep a short pointer there telling Codex to read CLAUDE.md; fallback files do not stack with AGENTS.md.
• To expose project .claude/skills to Codex, symlink .agents/skills -> ../.claude/skills instead of copying skill files.

Troubleshooting:

• kronael missing from Codex: run codex plugin marketplace upgrade kronael (or kronael-local for older installs), then codex plugin add kronael@kronael.
• Kronael skills missing in Codex after install: run the bridge prompt above, then start a new Codex thread and open /skills.
• Claude hooks missing after install: rerun $kronael-install; it merges hook wiring from settings-recommended.json.
• Codex says Skipped loading ... invalid SKILL.md: run make skills-frontmatter-fix, reinstall/bridge with $kronael-install, then start a new Codex thread. The lint extracts SKILL.md frontmatter, validates it with PyYAML, and fixes known loose forms Claude Code tolerated.

What's in the bundle

• Skills (skills/) — auto-activating context plus workflow commands (/commit, /ship, /refine, /diary, ...). The create router skill bundles the creative-output generators ported from NousResearch/hermes-agent; only ones that run locally are bundled. Index and rationale: skills/README.md.
• Agents (agents/) — task workers (@distill, @improve, @learn, @readme, @refine, @visual), mostly launched via slash commands.
• Hooks (hooks/) — lifecycle scripts: keyword nudging, LOCAL.md injection, rule re-injection across compaction, stop-time checks. Wiring lives in settings-recommended.json; see hooks/README.md.
• The global skill — development wisdom installed as ~/.claude/CLAUDE.md.