Add agent trajectory evaluation against declarative rubrics

[JE-Chen]

Agent-QA batch — deterministic scoring of an agent run. Full layers + tests + EN/Zh v36 docs + README.

Feature (utils/trajectory_eval, pure-stdlib)

• evaluate_trajectory(trajectory, rubric): a trajectory is the ordered list of {action, args, observation} steps a run took; a rubric is plain data — required_actions (+ ordered for relative order), forbidden_actions, max_steps, success_contains. Returns {passed, score, steps, checks}; score is the fraction of applicable checks passed, passed requires all, and each check is {name, passed, detail} so a failure pinpoints the violated expectation. An empty rubric trivially passes.
• The rubric is JSON-friendly, so it lives in JSON action files and travels over MCP. Executor AC_evaluate_trajectory (coerces JSON-string args from the visual builder); MCP ac_evaluate_trajectory; Builder under Agent.

Verification

• 11 tests pass (empty rubric, required present/missing, ordered ok/bad, forbidden, max_steps, success_contains, partial score 0.5, executor round-trip with JSON strings, wiring, facade); ruff clean; radon no CC≥C; bandit clean; PySide6-free.
• Rebased onto dev with the approval-testing (Add approval testing: verify artifacts against approved baselines #243) registrations; kept both.

[codacy-production]

Up to standards ✅

🟢 Issues 0 issues

Results:
0 new issues

View in Codacy

🟢 Metrics 45 complexity · -39 duplication

Metric	Results
Complexity	45
Duplication	-39

View in Codacy

_{NEW Get contextual insights on your PRs based on Codacy's metrics, along with PR and Jira context, without leaving GitHub. Enable AI reviewer
TIP This summary will be updated as you push new changes.}

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud