---
name: spielos-review-output
description: Evaluate an actual SpielOS run, review its pinned artifacts and record a separately authorized exact-action decision.
---

Select the company and inspect the run and its pinned rubric. Apply each stable
criterion with its evaluator and rating anchors. Record pass, fail, missing,
error or not-applicable honestly. Missing criteria and empty populations cannot
pass; a high average cannot hide a blocking failure. Label self-judging.

For images, pages or video, inspect the rendered artifact at relevant sizes;
file existence and a producer's description do not prove visual quality. Model
judgments need their pinned model, prompt, version and human calibration evidence.
Record artifact digests, observed evidence and reviewer identity through the tools.

Use the authenticated independent reviewer for `review_output` action `review`.
The same producer cannot substitute another reviewer name. Technical acceptance
and visual quality do not establish revenue or another unmeasured business outcome.

Only call `approve_action` after the owner authorizes that exact action and scope.
Bind its destination and content digest. This tool records the decision and does
not publish. Keep rejected, unknown and incomplete results visible.

Use `review_output` action `calibrate-judge` to register independent ratings of actual delivered artifact runs for a pinned judge. Do not submit invented reviewer names or rate the producer's own artifacts as independent evidence. When revising an evaluation, retain the same pinned artifact/code and explain the observed correction; prior reports remain in the event history.

Use action `brand-release` for independent brand review with the recorded candidate and inspection. The inspection's PNG must be one of that candidate's pinned example files. A producer cannot release it using another reviewer name. Historical brand references or a neutral starter never count as approval of an unseen new creative output.


Evaluation and observability are governed by `EVALUATION.md` and `spielos-evaluation/2`. Codex/Claude/OpenCode is the host; Director is a recorded SpielOS role in that host. Start or reuse the activity session supplied by the context hook; pass its session identity with runtime operations. If capture is missing, say what is unobserved and start explicit activity recording before claiming observable work. Log actual supplied/changed context, tool outcomes, measured artifacts and feedback targets. Classify each correction as run/project/owner/workflow/evaluation/build scope and state its recorded consequence. During an order, keep artifact revisions and bounded candidate Improvements distinct; never silently mutate the active release. Evals must judge actual behavior/output against explicit questions and anchored criteria. A filename, schema pass or successful command is only supporting evidence. Inspectable Langfuse inputs/outputs, datasets, experiment items, ratings and rationale must be confirmed remotely before qualification. Local passing evidence remains awaiting telemetry. Templates, synthetic tests and self-ratings never certify Director/model/creative quality. Record exposed usage with its source; hidden context/reasoning and unavailable host tokens stay unknown.


Use the selected company's live `/report` record for workflow/run counts, memory, supplied context, artifacts, evaluations and telemetry status. Read `company/evaluation-and-learning-contract.md` when present. Every new or changed workflow needs the executable evaluation pair and regression cases before qualification. Record actual host token metadata and separately sourced reference prices; never invent usage, billed cost, memory retention or independent acceptance. End meaningful work with a concise human tracking receipt and report/trace links. Keep company content in the selected home and reusable product code in source; no publication, hidden pursuit or cross-company routing is implied.

For setup or a missing company, use the bundled setup skill. Resolve the launcher at scripts/entry.py relative to the installed plugin root. Use it for report and experiment operations that are not exposed by the six MCP tools; never require a separate global SpielOS installation. The runtime operating guide is bundled at runtime/company/README.md. Keep those commands internal and give the user names, findings and next decisions.