model-setup
Skill2 starsChanged 7 days agoApache-2.0View on GitHub
Configure consumer-owned opencode model overrides for the cheap, standard, and deep QA lanes without pinning blueprint source or global config. Use for "/model-setup", "Modelle für QA-Lanes einrichten", or "configure QA models". Under Claude Code, explain the haiku/sonnet/opus mapping only and write no files.
In this repository34 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- accessibility-testingSkill
- agentswitchSkill
- api-testingSkill
- bug-reportSkill
- bug-reproductionSkill
- demoSkill
- deploy-to-stageSkill
- exploratory-testingSkill
- failure-analysisSkill
- onboard-productSkill
- onboard-projectSkill
- perf-analysisSkill
- qa-helpSkill
- qa-metricsSkill
- qa-reportSkill
- re-boardSkill
- reconSkill
- release-readinessSkill
- report-to-jiraSkill
- retest-bugSkill
- retro-learnSkill
- risk-based-testingSkill
- scaffold-suiteSkill
- security-testingSkill
- skill-reviewSkill
- test-automationSkill
- test-case-managementSkill
- test-reliabilitySkill
- test-ticketSkill
- visual-testingSkill
- webdriverio-testingSkill
- whats-newSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 26
Use the actual session directory as the consumer.
- Very specific: line 43
Add --cheap-variant <v>, --standard-variant <v>, or --deep-variant <v> only for user-selected supported variants.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
/model-setup — consumer QA model lanes
Identify the active harness first. The routing policy is:
| Route | Lane |
| Fast path / low risk | cheap |
| Complex path / low or medium risk | standard |
| High risk / escalated | deep |
| Direct path / no risk | standard |
Claude Code
Under Claude Code, only explain the mapping: cheap = haiku, standard = sonnet, deep = opus. Apply the routing table above with those names. Do not run the helper and do not create or edit any configuration file.
opencode
- Use the actual session directory as the consumer. The only config target is
<session>/opencode.json. Never edit the blueprint checkout, itsopencode/agents/*.md,opencode/opencode.template.jsonc, global opencode config, or top-levelmodel/small_modelchoices. The helper preserves unrelated config, MCP, permission, and model keys and backs up an existing consumer file toopencode.json.bak. - Let the user choose three model IDs and any supported variants from
opencode models. Never claim a model or variant is available from memory or assumption. If a variant is unknown or not shown as supported, recommend omitting it; do not write it blindly. Show the routing table and confirm the selected cheap, standard, and deep assignments. Use the session cwd explicitly:
python3 .claude/scripts/model-setup.py --cwd <session> --cheap <provider/model> --standard <provider/model> --deep <provider/model>Add
--cheap-variant <v>,--standard-variant <v>, or--deep-variant <v>only for user-selected supported variants. An omitted or empty variant removes that lane's existing variant. Use--dry-runfor a write-free preview. Model IDs useprovider/model; provider-qualified model IDs may contain further slashes in the model portion.The command sets only these consumer
agentoverrides:test-executor-cheap,test-executor,test-executor-deep,test-automator-cheap,test-automator, andtest-automator-deep. Inspect them read-only with:python3 .claude/scripts/model-setup.py --cwd <session> --check- Tell the user to restart opencode so the overrides take effect.
- Optionally offer a small
opencode run --model <provider/model> "Reply with OK"smoke test for a selected model. Run it only after the user explicitly confirms the smoke run and its model; never run it silently. A smoke run is optional verification, not proof that an unlisted variant is supported.
The helper uses Python's standard library only. It accepts JSON and JSONC with comments, escaped strings, and trailing commas. It performs no network, provider, or subprocess calls and never prints unrelated config values or secrets.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

