Skip to content

CowboyLogic/ai-dev / skills/client-config-opencode

client-config-opencode

Verified

Manage OpenCode configuration files (V2 native format by default, V1 supported) including MCP servers, providers, models, agents, commands, permissions, policies, skills, plugins, CLI/TUI settings, keybinds, and themes. Use this skill whenever the user wants to view, edit, add, migrate, or understand any opencode setting in opencode.json(c), cli.json, or tui.json. Trigger on: "configure opencode", "add provider", "set model", "add MCP server", "create agent", "configure permissions", "change theme", "add keybind", "show opencode config", "add custom command", "migrate opencode config to V2", or "opencode cli.json".

FileKindSize
CHANGELOG.mdmarkdown18.6 KB
README.mdmarkdown3.9 KB
SKILL.mdskill md8.8 KB
evals/evals.jsonconfig6.7 KB
references/agents.mdmarkdown5.4 KB
references/config-schema.mdmarkdown14.1 KB
references/mcp.mdmarkdown5.1 KB
references/permissions.mdmarkdown5.2 KB
references/providers.mdmarkdown10.8 KB
references/tui.mdmarkdown12.7 KB
references/v2/agents.mdmarkdown5.1 KB
references/v2/cli.mdmarkdown11.1 KB
references/v2/config-schema.mdmarkdown31.8 KB
references/v2/mcp.mdmarkdown10.3 KB
references/v2/migration.mdmarkdown8.9 KB
references/v2/permissions.mdmarkdown16.1 KB
references/v2/plugins.mdmarkdown8.6 KB
references/v2/providers.mdmarkdown21.4 KB
scripts/show-config.pyscript14.2 KB
scripts/update-references.pyscript3.9 KB
sources.jsonconfig7.1 KB
# OpenCode Configuration Manager

Evaluation results for the `client-config-opencode` skill. The skill itself is
[SKILL.md](SKILL.md), and its history is in [CHANGELOG.md](CHANGELOG.md).

> [!NOTE]
> **OpenCode version:** references verified against OpenCode V2
> [2.3.0](https://github.com/anomalyco/opencode/tree/v2.3.0) (tagged 2026-11-13) and V1
> [1.19.5](https://github.com/anomalyco/opencode/tree/v1.19.5) (tagged 2026-10-28),
> checked 2026-11-15. Neither binary was run for this check: claims that depend on a binary
> (`--help` output, `{file:...}` substitution) keep their 2.0.27 basis, and everything else comes
> from the tagged docs and source and the published `cli.json` schema. V1 changes since 1.19.0 rest on the tag
> diff and commit log because the V1 release notes were not reachable. V2 ships as git tags and
> binaries rather than GitHub Releases, so its version is the tag; its release notes are the repository's
> [CHANGELOG.md](https://github.com/anomalyco/opencode/blob/dev/CHANGELOG.md). Re-verify when either
> line ships a new release; what changed in this skill is in [CHANGELOG.md](CHANGELOG.md).

## Evaluation results

Benchmark run: **2026-04-26** · 4 scenarios · 8 total runs

These results predate the 2026-09-24 v2.0 refresh, which pinned these cases to V1 in
[evals/evals.json](evals/evals.json) and added three V2 cases.

### Overall

| | With skill | Baseline (no skill) | Delta |
|---|---|---|---|
| Assertions passed | 24 / 24 | 16 / 24 | — |
| Pass rate | **100%** | **67%** | **+33 pp** |

### By scenario

| Scenario | With skill | Baseline | Delta | Notes |
|---|---|---|---|---|
| Remote MCP server with OAuth + timeout (`sentry`, `type: remote`, `oauth` omit for auto, `timeout: 10000`) | 6 / 6 | 5 / 6 | **+17 pp** | Baseline set `timeout: 10` (seconds) instead of `10000` (ms); also used a non-existent `auth.type` schema for OAuth |
| GitLab Duo self-hosted provider (`GITLAB_INSTANCE_URL` env var, `{env:}` syntax, `provider` key) | 5 / 5 | 2 / 5 | **+60 pp** | Baseline used wrong top-level key (`providers` not `provider`), wrong env syntax (`env:VAR` not `{env:VAR}`), and embedded `apiBase` in config instead of using `GITLAB_INSTANCE_URL` env var |
| Custom read-only subagent (JSON + markdown format, `permission` schema, `mode`, temperature) | 8 / 8 | 4 / 8 | **+50 pp** | Baseline fabricated an entirely wrong schema: `agents` (plural), `type: subagent` instead of `mode: subagent`, and `tools.x: bool` instead of `permission.x: allow/deny` |
| Keybind leader key + disable (`tui.json`, `keybinds.leader`, empty string to disable) | 5 / 5 | 5 / 5 | 0 pp | Non-discriminating — TUI keybind config is generic enough that both configurations scored perfectly |

### Key takeaway

The baseline fails most on opencode-specific schema details that cannot be inferred from
general training data. Three patterns account for the +33 pp gap:

- **Permission model** — opencode uses `permission.x: "allow"/"deny"` (string values, singular
  `permission` key); the baseline consistently invents `tools.x: bool` or similar.
- **Provider config schema** — `{env:VAR}` substitution syntax, the singular `provider` key,
  and `GITLAB_INSTANCE_URL` as an env var (not an `apiBase` config field) are all
  opencode-specific and get wrong without the reference.
- **MCP transport details** — `timeout` is in milliseconds; the baseline treats it as seconds.

The one area where the skill adds no value in these evals is TUI keybind configuration —
both configurations scored perfectly on the keybind scenario, suggesting that test is
non-discriminating and should be replaced in a future iteration.

> [!TIP]
> **When does this skill matter most?** Any task touching the `permission` schema, provider
> configuration (especially newer providers like GitLab Duo, Helicone, llama.cpp), MCP
> transport options, or schema fields added after the model's training cutoff (LSP config,
> `tool_output`, `compaction.tail_turns`, command template syntax).
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 468kUpdated Claude CodeCodexUnverifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 319kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 319kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 319kUpdated Claude CodeCodexVerifiedInstall