agentleFS
Sign inStart free
agentstationagentstation/skills/model-selection/SKILL.md

model-selection

Skill1 starsChanged 2 months agoMITView on GitHub

Select or validate models and effort levels for delegated work, review, coordination, or focused specialist problems. Use when changing routing, refreshing benchmark evidence, or diagnosing model-policy conflicts. Do not use for ordinary work under an already valid assignment.

16 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use architectural-proposals, ci-validation-gates and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • Microsoft AzureSkill

    architectural-proposals

    How to write comprehensive architectural proposals that drive alignment before code is written

    Also in 3 projects that use this

  • Microsoft AzureSkill

    ci-validation-gates

    Defensive CI/CD patterns: semver validation, token checks, retry logic, draft detection — earned from v0.8.22

    Also in 3 projects that use this

  • +10 more from Azure

The file

We split this skill into the rules it gives an agent

12 rules. 0 are common, 2 are distinctive, and 10 are very specific to this project.

  • 0

    Common

    Most of the 13 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 10

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the host's actual model and effort catalog.

    line 18
  2. Very specific:

    Read references/selection.md for availability input and role evidence.

    line 25
  3. Very specific:

    Read references/refresh.md before changing the catalog or policy.

    line 43
  4. Very specific:

    Keep source dates, benchmark versions, unrounded measurements, and uncertainty.

    line 44
  5. Very specific:

    Use scripts/model-selection frontier to recompute the frontier.

    line 45
  6. Very specific:

    Use scripts/model-selection table --check PATH to verify the global instructions.

    line 46
  7. Distinctive:

    Do not substitute vendor claims for measured results.

    line 47
  8. Very specific:

    Import scripts/model_policy.py from the canonical installed sibling skill.

    line 51
  9. Very specific:

    Call validate_assignment for the primary model and every fallback.

    line 52
  10. Very specific:

    Read references/integration.md for the API and migration contract.

    line 55
  11. Very specific:

    Read references/benchmark-evidence.md only for a benchmark evidence question.

    line 57
  12. Very specific:

    Run python3 -m unittest discover -s model-selection/tests -v from the collection root after a change.

    line 58

“Common” is measured: kept by at least 6 in 10 of the 13 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Model selection

This skill owns model eligibility and benchmark evidence across projects. The calling host owns dispatch and its available model catalog. Global instructions retain the owner's visible model tables and role restrictions.

Select an assignment

  1. Identify the task class, role, complexity, and quality requirements.
  2. Read the host's actual model and effort catalog.
  3. Validate a proposed assignment with scripts/model-selection check.
  4. For a comparison, pass host availability to scripts/model-selection select.
  5. Give the validated assignment to the native dispatch adapter.

The checker does not invoke a model or change the current session. A successful policy check does not prove host availability. Read references/selection.md for availability input and role evidence.

export MODEL_SELECTION="${AGENTS_HOME:-$HOME/.agents}/skills/model-selection/scripts/model-selection"
"$MODEL_SELECTION" check \
  --model gpt-6.1-sol --effort high --role worker
"$MODEL_SELECTION" select \
  --task implementation --available-json host-models.json

The role policy applies before explicit profiles, ranking, and fallback. Fable only coordinates work. Do not use Haiku. Astra requires a narrow, highly complex, focused specialist assignment. Opus coordination or high-complexity work requires multi-agent Ultracode. The caller must verify those task conditions before setting the corresponding flags.

Refresh evidence

Read references/refresh.md before changing the catalog or policy. Keep source dates, benchmark versions, unrounded measurements, and uncertainty. Use scripts/model-selection frontier to recompute the frontier. Use scripts/model-selection table --check PATH to verify the global instructions. Do not substitute vendor claims for measured results.

Integrate a caller

Import scripts/model_policy.py from the canonical installed sibling skill. Call validate_assignment for the primary model and every fallback. Missing policy data stops dispatch with an installation recovery message. A CLI override cannot bypass a role restriction. Read references/integration.md for the API and migration contract.

Read references/benchmark-evidence.md only for a benchmark evidence question. Run python3 -m unittest discover -s model-selection/tests -v from the collection root after a change.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.

Check your setup, free