run

Part of the ACED plugin — see that page for install instructions.

Trigger: run after editing a skill, AGENTS.md section, subagent, or command, to score its current behavior against its golden set

Runs the frozen .feature suite for a target agent configuration and reports pass rate and per-layer results. The .feature is the single eval source — boolean scenarios, inline @rubric criteria, and the @trigger Scenario Outline all live in it; eval.md carries only the subject binding and run policy.

What it does

1. Locates the target’s node in the project spec — .agents/specs/<project>/…/<node>/, discovered through the SDD spec tree (the node’s eval.md names the subject) — which holds the frozen <node>.feature and its colocated eval.md (asks if multiple nodes carry an eval.md; stops and reports if no suite is initialized).
2. Reads the run policy from eval.md — subject, eval.layers, eval.judge.{model, default_threshold}, and eval.trigger.{activation_threshold, runs} — then reads the current subject agent configuration in full.
3. For each scenario in the frozen .feature, invokes the aced-case-judge subagent blind — passing only the .feature path and the scenario name, never the Then or the rubric — skipping any layer not enabled in eval.md. A @rubric scenario carries its rubric inline; a @trigger Scenario Outline runs one case per Examples row.
4. Runs every scenario before reporting (does not stop on first failure).
5. Writes results to .agents/aced/results/<target-slug>/<ISO8601-timestamp>.json — the shared, git-ignored ACED results directory at the repo root, keyed by the target — recording each scenario’s per-dimension scores, total against its own maximum, threshold, and pass/fail.
6. Reports pass rate, per-layer breakdown, and failing scenarios sorted worst-first by margin.

Next step

If cases fail, run improve to diagnose the pattern. If pass rate is 100%, run add-scenario to expand edge-case coverage.