Menú
How to delegate a task to an opencode pseudo-subagent (default model Kimi K2.6). Use ONLY when the user explicitly invokes /opencode or names opencode by name — not for general subagent or delegation requests.
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Basado en la clasificación ocupacional SOC
Debugging a ghostty crash
Drive a real browser during frontend work — verify a change rendered, click through a flow, fill forms, read console errors, take screenshots, or debug performance/memory/network. Use whenever interacting with a local web app in a browser (Playwright script templates, chrome-devtools tradeoffs, SPA mock-API gotchas).
How to use the tviz CLI to read and write Things 3 tasks
For non-Codex agents, get a Codex second opinion on your changes. Claude sends the diff to Codex for review, discusses the feedback, and synthesizes actionable results.
Productivity coach
Read Oxide RFD contents, metadata, and images by number using rfd-cli. Use when an RFD comes up (e.g. "RFD 63", "what does the networking RFD say") and you need the actual text.
| name | opencode |
| description | How to delegate a task to an opencode pseudo-subagent (default model Kimi K2.6). Use ONLY when the user explicitly invokes /opencode or names opencode by name — not for general subagent or delegation requests. |
opencode run is the claude -p equivalent: a one-shot prompt to a full agent
(file/bash/grep tools, runs in cwd). Use it to delegate self-contained tasks to a
cheaper model. Default model is always opencode/kimi-k2.6 (OpenCode Zen-hosted,
cheap and capable) unless the user specifies another.
opencode run -m opencode/kimi-k2.6 --dangerously-skip-permissions "<task>"
--dangerously-skip-permissions is required: the run is non-interactive, so any
ask permission would hang forever waiting for approval.--agent) is the full tier: read, edit, bash, everything.Restricted tiers are named agents that pin the model to kimi-k2.6 and remove tools.
Select with --agent:
| Tier | Invoke | Can | Cannot |
|---|---|---|---|
| full | (no --agent) | read, edit, bash, spawn subagents | — |
| edit | --agent oc-edit | read, search, edit files, webfetch | run shell, spawn subagents |
| explore | --agent oc-explore | read, search, browse the web | edit, run shell, spawn subagents |
opencode run --agent oc-explore --dangerously-skip-permissions "<read-only task>"
opencode run --agent oc-edit --dangerously-skip-permissions "<edit task, no shell>"
The tier files live in opencode/agent/*.md in the dotfiles repo (symlinked to
~/.config/opencode/agent/ by install.sh). To add/change a tier, edit those and
rerun install.sh. Note: locking down a tier requires denying task too — otherwise
the agent escapes its sandbox by spawning a default subagent that has the tools back.
Override the model with -m <provider>/<model> (e.g. -m google/…, -m openai/…).
opencode models lists options; opencode auth list shows authed providers. -m
overrides a tier's baked-in model.
> <agent> · <model> banner. For machine-readable
output add --format json.