CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

choose-model

Recommend a currently supported model capability and effort for a described task, considering uncertainty, error cost, scope and latency. Use when the user asks which model or reasoning level to use, or needs routing advice for authorised subagents. Verify current availability before naming a model or control.

Quality

81%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Choose Model

Recommend an adequate supported capability for the task. Keep the user's main-chat selection unchanged. A recommendation is not a setting change or evidence of execution.

Establish current choices

Inspect the current host's exposed model inventory and tool schema, or current authoritative product documentation when access is authorised. Name an exact model, argument or effort value only when that evidence supports it. Do not assume a copied model lineup, a named API skill, a built-in command or an Agent effort parameter exists on this host.

If availability cannot be verified, recommend a capability class and state that limitation. Continue suitable authorised work; do not use paid API workarounds, install missing skills or change credentials to reproduce unavailable routing. Respect applicable project allocation, independent review and explicitly selected specialist programme floors.

Weigh the task

  • Mechanical, well-specified, reversible work can use a fast, lower-cost supported route.
  • Routine implementation or analysis needs a capable coding/reasoning route.
  • Ambiguous architecture, concurrency, security, clinical or irreversible decisions need sufficient reasoning and required specialist review for their consequences.
  • Large context alone is not a reason to select the highest tier; reduce irrelevant input first.
  • Creative work may benefit from a supported capability suited to tone or narrative.
  • When adequate choices are tied, prefer the lower cost or latency. Do not claim unobserved prices or performance.

Use only the effort controls exposed by the relevant interface. Accept the user's ordinary-language confirmation; no special marker or resubmission is needed. Missing model/effort selection means routing cannot be enforced, not that an invented parameter should be supplied.

Output

Keep it short: recommendation, the deciding factor, and any availability limitation. Distinguish requested routing from actual execution telemetry; a tool accepting a request proves only that it accepted the request. Ask a brief question only if the task itself is too unclear to recommend an adequate capability.

Repository

BigSimmo/PsychSift

Path

.claude/cloud-profile/skills/choose-model/SKILL.md

Last updated
First committed

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.