Point skill-evals at skillscope's first-class commands

[danielholanda]

Pull Request Template

Note to AMDers:
This is a public repository. Please do not upload any confidential or customer data. Make sure all such data has been anonymized or removed before making this PR. If you need to attach any private files or links, please insert a Internal OneDrive Link or a Jira Ticket Link instead.

Updates .github/workflows/skill-evals.yml for the recent danielholanda/skillscope changes (#2–#4).

The three graders are commands now, not modes of run. The workflow stays the same three jobs, one checkout-plus-action each:

Job	Was	Now
structural	command: structural	unchanged
routing	command: run + --mode routing --routing-skills analysis-orchestrator	command: routing
behavioral	command: run + --mode behavior	command: behavioral

--routing-skills is gone because this repo has one skill: that skill is the room. The behavior job is renamed behavioral to match the command.

No dataset or skill-body change. evals.json already matches the current schema.

Verification

Locally, against danielholanda/skillscope@main:

$ skillscope run
error: argument command: invalid choice: 'run'
        (choose from 'structural', 'routing', 'behavioral', ...)

$ skillscope structural --skills TraceLens/Agent/Analysis/skills/analysis-orchestrator
[evals] OK: 1 skill folder(s) in shape.
[evals] OK: 12 case(s) across 1 skill(s) plus 5 shared negative(s).
[evals] OK: 56 internal reference(s) across 17 markdown file(s).

$ skillscope select --all --skills TraceLens/Agent/Analysis/skills/analysis-orchestrator
{"routing": true, "default": [{"skill": "analysis-orchestrator", ...}]}

CI on this PR:

• structural passed
• routing and behavioral launched the new commands (skillscope routing / skillscope behavioral) and then failed at the Claude preflight with Not logged in. Same failure as Run the analysis-orchestrator evals with skillscope #3: ANTHROPIC_API_KEY is not a repository secret, so those two jobs cannot grade yet.