Poll OSS lab completion and add screenshot fallback

docs/architecture/local-codex-tui-actor.md

@@ -0,0 +1,152 @@
+# Local Codex TUI Actor Contract
+
+Date: 2026-06-02
+
+Status: first implementation slice on issue #28. The repo now supports one
+explicit local actor with sanitized lifecycle evidence and fail-fast Codex
+workspace-trust preflight; four-lane fanout and trust bootstrap remain
+follow-up work.
+
+## Goal
+
+Let `mimetic run` dogfood `mimetic-cli` with a real local Codex TUI actor while
+preserving the public-safe harness contract:
+
+```text
+spawn actor
+-> submit bounded prompt
+-> stream sanitized lifecycle events
+-> write run bundle
+-> render Observer while active
+-> verify final bundle
+```
+
+This actor proves the local harness substrate. It must not claim target product
+behavior beyond the commands and artifacts it actually observes.
+
+## Explicit Opt-In
+
+The first live actor must require an explicit flag or env var. Non-dry-run
+without opt-in should continue to fail closed.
+
+Suggested gate:
+
+```bash
+mimetic run --actor codex-tui --sims 1 --timeout-ms 120000
+```
+
+or:
+
+```bash
+MIMETIC_ENABLE_LOCAL_CODEX_TUI=1 mimetic run --sims 1
+```
+
+Provider spend, E2B, GitHub mutation, and external network calls remain off
+unless separately and explicitly requested.
+
+For deterministic tests or local substrate debugging, `MIMETIC_CODEX_ACTOR_COMMAND`
+can point at a safe fixture command. Real Codex TUI launch uses a Linux
+`script` PTY wrapper when available because the Codex TUI requires a terminal.
+
+## Lifecycle Events
+
+The actor should append deterministic events to `events.ndjson`:
+
+| Event | Required fields |
+| --- | --- |
+| `actor.spawned` | `simId`, `streamId`, command name, cwd, startedAt |
+| `actor.prompt.submitted` | prompt digest, prompt class, no raw prompt if unsafe |
+| `actor.observation` | sanitized transcript tail, byte count, redaction status |
+| `actor.artifact` | relative artifact path, kind, digest |
+| `actor.verdict` | `passed`, `failed`, `blocked`, or `timed_out`; reason |
+| `actor.exited` | exit code, signal, durationMs |
+| `actor.timeout` | timeoutMs, last safe observation |
+| `actor.cancelled` | signal and operator reason |
+
+The Observer should be able to render a live terminal stream from those events
+while the actor is active, then render the final transcript and artifacts after
+exit.
+
+## Runtime State
+
+All generated state stays under ignored `.mimetic/`:
+
+```text
+.mimetic/runs/<run-id>/
+  run.json
+  review.json
+  review.md
+  events.ndjson
+  actor.json
+  transcripts/codex-tui-sanitized.txt
+  observer/
+    index.html
+    observer-data.json
+```
+
+No raw terminal output is public by default. The transcript artifact must be
+sanitized before it is linked from the bundle.
+
+## Redaction Rules
+
+Before any transcript tail or event payload is written:
+
+- redact OpenAI, E2B, GitHub, npm, and generic private-key patterns;
+- redact absolute local home/workspace paths when they are not necessary for
+  proof;
+- block the run if redaction cannot prove `status: passed`;
+- record env var names only, never values.
+
+## Initial Prompt Class
+
+The first actor prompt should be bounded to public-safe dogfood work:
+
+- inspect `mimetic/` dogfood config;
+- run `pnpm mimetic -- doctor`;
+- run or explain the strongest safe Mimetic proof command available;
+- do not commit, push, publish, file issues, or print secrets;
+- summarize blockers using public-safe evidence paths.
+
+The raw prompt can live in source only if it contains no private context and no
+credential values.
+
+## Stop Conditions
+
+The actor must stop and mark the lane `blocked` or `failed` if:
+
+- Codex CLI is not installed or not authenticated;
+- a command requests approval in a non-interactive run;
+- redaction fails;
+- output contains a likely secret after redaction;
+- the actor touches files outside allowed runtime paths without explicit scope;
+- timeout is reached.
+
+## Known First-Slice Boundary
+
+The first real local TUI proof in this environment reached Codex's workspace
+trust prompt. The current implementation now detects that state before spawn
+and writes a `blocked` run bundle instead of waiting for the TUI timeout. The
+resulting bundle is still verifiable and Observer-renderable, but it does not
+prove autonomous actor completion until the trust root is explicitly approved.
+
+The next implementation slice should add an explicit, public-safe trust
+bootstrap before spawning the TUI. Acceptable fixes include:
+
+- allow an explicit operator-approved trust bootstrap for this repository only;
+- or add a non-TUI `codex exec` actor mode as a separate contract while keeping
+  the TUI actor honest about needing a PTY and trust.
+
+## Acceptance For First Slice
+
+The implementation slice for this spec should prove:
+
+```bash
+pnpm mimetic:doctor
+pnpm mimetic -- run --actor codex-tui --sims 1 --timeout-ms 120000 --json
+pnpm mimetic -- watch --run latest --detach --json --no-open
+pnpm mimetic -- verify --run latest --json
+pnpm check
+```
+
+If the 1x actor works, split or follow with a 4x fanout issue. Do not start with
+4x fanout before the single actor lifecycle is deterministic.

src/observer-assets.ts

@@ -952,18 +952,22 @@ a { color: inherit; }
   background: var(--obs-blue);
 }
 
-.tile-cap[data-status="failed"] .tile-cap-bar > span,
-.tile-cap[data-status="blocked"] .tile-cap-bar > span { background: var(--obs-red); }
+.tile-cap[data-status="failed"] .tile-cap-bar > span { background: var(--obs-red); }
+.tile-cap[data-status="blocked"] .tile-cap-bar > span,
+.tile-cap[data-status="timed_out"] .tile-cap-bar > span { background: var(--obs-amber); }
 
+.tile-cap[data-status="passed"] .tile-cap-bar > span,
 .tile-cap[data-status="complete"] .tile-cap-bar > span,
 .tile-cap[data-status="contract_proof_only"] .tile-cap-bar > span { background: var(--obs-green); }
 
 .pip[data-status="queued"] { color: var(--obs-fg-4); }
 .pip[data-status="preparing"] { color: var(--obs-sky); }
 .pip[data-status="running"] { color: var(--obs-blue); }
+.pip[data-status="passed"] { color: var(--obs-green); }
 .pip[data-status="complete"] { color: var(--obs-green); }
 .pip[data-status="contract_proof_only"] { color: var(--obs-green); }
 .pip[data-status="blocked"] { color: var(--obs-amber); }
+.pip[data-status="timed_out"] { color: var(--obs-amber); }
 .pip[data-status="failed"] { color: var(--obs-red); }
 
 .focus {
@@ -1081,7 +1085,8 @@ a { color: inherit; }
   color: #8aa9ff;
 }
 
-.focus-status-badge[data-status="blocked"] {
+.focus-status-badge[data-status="blocked"],
+.focus-status-badge[data-status="timed_out"] {
   background: var(--obs-amber-soft);
   color: #e9b04b;
 }
@@ -1091,6 +1096,7 @@ a { color: inherit; }
   color: #ee8a7e;
 }
 
+.focus-status-badge[data-status="passed"],
 .focus-status-badge[data-status="complete"],
 .focus-status-badge[data-status="contract_proof_only"] {
   background: var(--obs-green-soft);
@@ -1619,7 +1625,7 @@ export function observerClientJs(): string {
     var surface = document.createElement("div");
     surface.className = "media-surface";
 
-    if (preferScreenshots && stream.ui && stream.ui.screenshotUrl) {
+    if (stream.ui && stream.ui.screenshotUrl && (preferScreenshots || (stream.embed && stream.embed.kind === "screenshot"))) {
       var img = document.createElement("img");
       img.src = stream.ui.screenshotUrl;
       img.alt = "Screenshot for " + stream.id;
@@ -1690,7 +1696,15 @@ export function observerClientJs(): string {
     spinner.setAttribute("aria-hidden", "true");
     var title = document.createElement("div");
     title.className = "live-waiting-title";
-    title.textContent = stream.status === "contract_proof_only" ? "Contract proof only" : "Waiting for live desktop";
+    title.textContent = stream.status === "contract_proof_only"
+      ? "Contract proof only"
+      : stream.status === "failed"
+        ? "Live desktop failed"
+        : stream.status === "timed_out"
+          ? "Live desktop timed out"
+          : stream.status === "blocked"
+            ? "Live desktop blocked"
+            : "Waiting for live desktop";
     var url = document.createElement("div");
     url.className = "live-waiting-url";
     url.textContent = stream.ui && stream.ui.route ? stream.ui.route : stream.label;
@@ -2071,8 +2085,8 @@ export function observerClientJs(): string {
       }
       if (activeStatus === "all") return true;
       if (activeStatus === "running") return stream.status === "running" || stream.status === "preparing";
-      if (activeStatus === "blocked") return stream.status === "blocked" || stream.status === "failed";
-      if (activeStatus === "complete") return stream.status === "complete";
+      if (activeStatus === "blocked") return stream.status === "blocked" || stream.status === "failed" || stream.status === "timed_out";
+      if (activeStatus === "complete") return stream.status === "complete" || stream.status === "passed";
       if (activeStatus === "proof") return stream.status === "contract_proof_only";
       return true;
     });
@@ -2081,9 +2095,9 @@ export function observerClientJs(): string {
   function countStatuses(streams) {
     return streams.reduce(function (acc, stream) {
       if (stream.status === "running" || stream.status === "preparing") acc.running += 1;
-      else if (stream.status === "blocked") acc.blocked += 1;
+      else if (stream.status === "blocked" || stream.status === "timed_out") acc.blocked += 1;
       else if (stream.status === "failed") acc.failed += 1;
-      else if (stream.status === "complete") acc.complete += 1;
+      else if (stream.status === "complete" || stream.status === "passed") acc.complete += 1;
       else if (stream.status === "contract_proof_only") acc.proof += 1;
       else acc.queued += 1;
       return acc;

src/observer-data.ts

@@ -15,7 +15,7 @@ export interface ObserverData {
   run: {
     runId: string;
     mode: RunBundle["mode"];
-    status: "contract_proof_only" | "running" | "complete" | "failed";
+    status: RunBundle["review"]["verdict"];
     title: string;
     createdAt: string;
     simCount: number;
@@ -74,7 +74,7 @@ export function buildObserverData(bundle: RunBundle, generatedAt = new Date().to
   });
 
   const warnings = events.filter((event) => event.level === "warn").length;
-  const blocked = streams.filter((stream) => stream.status === "blocked" || stream.status === "failed").length;
+  const blocked = streams.filter((stream) => stream.status === "blocked" || stream.status === "failed" || stream.status === "timed_out").length;
   const active = streams.filter((stream) => stream.status === "running" || stream.status === "preparing").length;
 
   return {
@@ -175,10 +175,14 @@ function statusLabel(status: RunStream["status"]): string {
       return "Queued";
     case "running":
       return "Running";
+    case "passed":
+      return "Passed";
     case "complete":
       return "Complete";
     case "blocked":
       return "Blocked";
+    case "timed_out":
+      return "Timed out";
     case "failed":
       return "Failed";
   }