humanish 0.55.0 → 0.56.0

package/dist/tui-contract.d.ts

@@ -90,3 +90,20 @@ export declare function nodeSupportsTui(versionString?: string): boolean;
  * readiness row that reports it can never look in different places.
  */
 export declare function tuiBundleUrl(baseUrl: string): URL;
+/**
+ * What `doctor` says about the stakeholder surface — a pure function of the machine's state and,
+ * crucially, of WHO IS READING.
+ *
+ * A real first-contact study (labs/first-contact.yaml) is why the reader matters. An autonomous
+ * agent evaluating humanish read "`humanish tui` is available in an interactive terminal",
+ * correctly concluded it was not in one, and dropped it — then wrote a report FOR A HUMAN that
+ * never mentioned the human surface at all. Discovery worked and handoff did not. A capability
+ * described to a reader who cannot use it has to be phrased as something to PASS ON, or it reads
+ * as "not for you" and dies in the transcript.
+ */
+export declare function terminalSurfaceMessage(state: {
+    supported: boolean;
+    bundlePresent: boolean;
+    interactive: boolean;
+    nodeVersion: string;
+}): string;

package/dist/tui-contract.js

@@ -29,4 +29,26 @@ export function nodeSupportsTui(versionString = process.version) {
 export function tuiBundleUrl(baseUrl) {
     return new URL("./tui-app.js", baseUrl);
 }
+/**
+ * What `doctor` says about the stakeholder surface — a pure function of the machine's state and,
+ * crucially, of WHO IS READING.
+ *
+ * A real first-contact study (labs/first-contact.yaml) is why the reader matters. An autonomous
+ * agent evaluating humanish read "`humanish tui` is available in an interactive terminal",
+ * correctly concluded it was not in one, and dropped it — then wrote a report FOR A HUMAN that
+ * never mentioned the human surface at all. Discovery worked and handoff did not. A capability
+ * described to a reader who cannot use it has to be phrased as something to PASS ON, or it reads
+ * as "not for you" and dies in the transcript.
+ */
+export function terminalSurfaceMessage(state) {
+    if (!state.supported) {
+        return `\`humanish tui\` needs Node ${TUI_MIN_NODE_MAJOR}+ (this is ${state.nodeVersion}); every other command works here`;
+    }
+    if (!state.bundlePresent) {
+        return "`humanish tui` bundle is not built in this checkout — run `pnpm build` (installed packages ship it prebuilt)";
+    }
+    return state.interactive
+        ? "`humanish tui` opens the interactive surface for browsing labs and runs"
+        : "`humanish tui` is the interactive surface a PERSON uses to watch and start runs here — this shell is not a terminal, so pass it on to whoever is";
+}
 //# sourceMappingURL=tui-contract.js.map

package/dist/tui-contract.js.map

@@ -1 +1 @@
-{"version":3,"file":"tui-contract.js","sourceRoot":"","sources":["../src/tui-contract.ts"],"names":[],"mappings":"AAAA,2DAA2D;AAC3D,EAAE;AACF,6FAA6F;AAC7F,mGAAmG;AACnG,+EAA+E;AAC/E,EAAE;AACF,gGAAgG;AAChG,+FAA+F;AAC/F,gGAAgG;AAChG,mGAAmG;AACnG,6DAA6D;AAC7D,EAAE;AACF,mGAAmG;AACnG,mBAAmB;AAwFnB,mFAAmF;AACnF,MAAM,CAAC,MAAM,kBAAkB,GAAG,EAAE,CAAC;AAErC;;;GAGG;AACH,MAAM,UAAU,eAAe,CAAC,gBAAwB,OAAO,CAAC,OAAO;IACrE,MAAM,KAAK,GAAG,MAAM,CAAC,QAAQ,CAAC,aAAa,CAAC,OAAO,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,EAAE,EAAE,CAAC,CAAC;IACvF,OAAO,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC,IAAI,KAAK,IAAI,kBAAkB,CAAC;AAC/D,CAAC;AAED;;;GAGG;AACH,MAAM,UAAU,YAAY,CAAC,OAAe;IAC1C,OAAO,IAAI,GAAG,CAAC,cAAc,EAAE,OAAO,CAAC,CAAC;AAC1C,CAAC"}
+{"version":3,"file":"tui-contract.js","sourceRoot":"","sources":["../src/tui-contract.ts"],"names":[],"mappings":"AAAA,2DAA2D;AAC3D,EAAE;AACF,6FAA6F;AAC7F,mGAAmG;AACnG,+EAA+E;AAC/E,EAAE;AACF,gGAAgG;AAChG,+FAA+F;AAC/F,gGAAgG;AAChG,mGAAmG;AACnG,6DAA6D;AAC7D,EAAE;AACF,mGAAmG;AACnG,mBAAmB;AAwFnB,mFAAmF;AACnF,MAAM,CAAC,MAAM,kBAAkB,GAAG,EAAE,CAAC;AAErC;;;GAGG;AACH,MAAM,UAAU,eAAe,CAAC,gBAAwB,OAAO,CAAC,OAAO;IACrE,MAAM,KAAK,GAAG,MAAM,CAAC,QAAQ,CAAC,aAAa,CAAC,OAAO,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,EAAE,EAAE,CAAC,CAAC;IACvF,OAAO,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC,IAAI,KAAK,IAAI,kBAAkB,CAAC;AAC/D,CAAC;AAED;;;GAGG;AACH,MAAM,UAAU,YAAY,CAAC,OAAe;IAC1C,OAAO,IAAI,GAAG,CAAC,cAAc,EAAE,OAAO,CAAC,CAAC;AAC1C,CAAC;AAED;;;;;;;;;;GAUG;AACH,MAAM,UAAU,sBAAsB,CAAC,KAKtC;IACC,IAAI,CAAC,KAAK,CAAC,SAAS,EAAE,CAAC;QACrB,OAAO,+BAA+B,kBAAkB,cAAc,KAAK,CAAC,WAAW,mCAAmC,CAAC;IAC7H,CAAC;IACD,IAAI,CAAC,KAAK,CAAC,aAAa,EAAE,CAAC;QACzB,OAAO,8GAA8G,CAAC;IACxH,CAAC;IACD,OAAO,KAAK,CAAC,WAAW;QACtB,CAAC,CAAC,yEAAyE;QAC3E,CAAC,CAAC,kJAAkJ,CAAC;AACzJ,CAAC"}

package/docs/contracts/schemas.md

@@ -3,7 +3,7 @@
 Date: 2026-06-02 (current-state note updated 2026-07-14)
 
 Status: reference map for the major contracts shipped through source version
-`0.55.0`; it is not an exhaustive inventory of command/result envelopes. Exported types,
+`0.56.0`; it is not an exhaustive inventory of command/result envelopes. Exported types,
 schema constants, parsers, and validators in `src/` are authoritative. Rows
 marked "reserved" name layering intent only — no code emits or validates them
 yet. Do not emit a reserved schema.
@@ -749,9 +749,10 @@ Core-owned fields:
   value is what produced the trace, so reporting it is honest where reporting
   "unset" would not be. Absent when a provider declares no settings and on every
   pre-existing bundle; tolerated by verify. It exists because effort was
-  unreachable from a lab, which made every run take the default silently, and a
-  trace that cannot say what effort produced it cannot be compared with one that
-  can
+  unreachable from a lab, which made every run take the default silently. Effort
+  is part of WHO the participant was rather than of how the instrument was tuned
+  (docs/principles/actor-fidelity.md), so a trace without it is a result missing
+  half its sample description
 - optional `affordanceUse` (`humanish.affordance-use.v1`): which KIND of route this
   actor took (see Affordance Use below)
 - optional `estimatedCost` (`humanish.actor-estimated-cost.v1`): a token-derived

package/docs/goals/current.md

@@ -29,7 +29,7 @@ study completed, reproduced, and produced a real accessibility finding via a
 keyboard-first participant
 ([docs/goals/email-gated-signup/receipts/](email-gated-signup/receipts/)).
 
-## Current Program Truth (source `0.55.0`)
+## Current Program Truth (source `0.56.0`)
 
 The package source and repository implementation in this tree agree on these
 points:
@@ -46,7 +46,7 @@ The immutable 2026-06-10 proof-roadmap packet is paired with a
 | Public proof | A legible four-persona Observer hero from a verified real public-application study (commit-pinned drawDB) shipped in the npm payload (`0.16.0`) | Coverage beyond a single studied subject; the stratified breadth panel remains unbuilt |
 | OSS meta-lab | Dry-run contract and separate disposable smoke harness | Live meta-lab execution; disabled until repository instructions and actor credentials have an isolated boundary |
 | Observer serving | `watch`/`observe` loopback servers plus `serve` — the run-library surface with loopback default, capability-link exposure, `share_ready`-gated open mode, and optional operator-run tunnel; streams never served remotely | A remote live-stream (`--live-streams`) design; a persistent capability-link store |
-| Stakeholder terminal surface | `humanish tui` (`0.50.0`, redesigned to the reviewed spec in `0.51.0`): labs -> lab -> run, arrow-key navigation, and starting a dry or live run from the lab screen. The run is DETACHED and outlives the terminal — live-proven by killing the terminal 42s into a real run that then ran on for ~3.5 minutes and finished `pass` at $0.639751. The run screen leads with the participant and their recorded thinking, live-proven mid-flight against a real computer-use run. Ships as one bundled file loaded on demand; refuses a non-interactive stdin/stdout naming the JSON commands instead. `0.51.0` builds the reviewed rev-8 design: wordmark and project context, content capped at 96 columns, live rows naming the PARTICIPANT and their elapsed clock, spinners and verdict glyphs, breadcrumbs, the lab's subject/model/caps/keys line, and one Start with a dry-run/live toggle. `0.52.0` completes the reviewed screen set: the run outcome card (denominator first, the participant's closing words, then Open in Observer / Run again), the interrupted card with Reclaim — live-proven by killing a real run and stopping its orphaned sandbox — and All runs, the cross-lab peer where participants lead and one thought line follows the cursor. `0.53.0` gives it the reviewed palette rather than the terminal's theme: exact hex on a truecolor terminal, downsampled where not, and never colour alone. `0.54.0` closes the last two gaps: stopping a running run (armed, signals the process GROUP, and says plainly that sandboxes are separate), and pricing a run WHILE it runs — the running usage now travels with the trace into the mid-run flush | Cancelling a run from the surface; per-user persisted config (#470); export (#471); aggregate stats (#472); agent-authored labs (#473); a reusable persona panel (#474) |
+| Stakeholder terminal surface | `humanish tui` (`0.50.0`, redesigned to the reviewed spec in `0.51.0`, reworked again in `0.56.0` from stakeholder feedback: two explicit start rows instead of a hidden mode toggle, a description line so a list of studies says what they are, `?` keys, and ←/→ back to meaning back and open): labs -> lab -> run, arrow-key navigation, and starting a dry or live run from the lab screen. The run is DETACHED and outlives the terminal — live-proven by killing the terminal 42s into a real run that then ran on for ~3.5 minutes and finished `pass` at $0.639751. The run screen leads with the participant and their recorded thinking, live-proven mid-flight against a real computer-use run. Ships as one bundled file loaded on demand; refuses a non-interactive stdin/stdout naming the JSON commands instead. `0.51.0` builds the reviewed rev-8 design: wordmark and project context, content capped at 96 columns, live rows naming the PARTICIPANT and their elapsed clock, spinners and verdict glyphs, breadcrumbs, the lab's subject/model/caps/keys line, and one Start with a dry-run/live toggle. `0.52.0` completes the reviewed screen set: the run outcome card (denominator first, the participant's closing words, then Open in Observer / Run again), the interrupted card with Reclaim — live-proven by killing a real run and stopping its orphaned sandbox — and All runs, the cross-lab peer where participants lead and one thought line follows the cursor. `0.53.0` gives it the reviewed palette rather than the terminal's theme: exact hex on a truecolor terminal, downsampled where not, and never colour alone. `0.54.0` closes the last two gaps: stopping a running run (armed, signals the process GROUP, and says plainly that sandboxes are separate), and pricing a run WHILE it runs — the running usage now travels with the trace into the mid-run flush | Cancelling a run from the surface; per-user persisted config (#470); export (#471); aggregate stats (#472); agent-authored labs (#473); a reusable persona panel (#474) |
 | Off-app comms | Vendor-neutral in-sandbox email/SMS catch, a minimal persona inbox surface, and digest-only `humanish.comms-thread.v1` evidence; wired into the computer-use and shared-world routes over both HTTP and SMTP; live-proven end to end on 2026-08-08 — a persona signed up for a public app, read the emailed link in its inbox, and reached the signed-in product (`docs/goals/email-gated-signup/receipts/signup-verify-live-2026-08-08.md`); the adopter-hosted / app-url ingress plane is wired on the CUA and concurrent external-public routes (#387/#380, 2026-08-11) | Real-provider delivery; a live adopter-hosted receipt |
 
 Capability proof and adopter replacement are different gates. A deterministic

package/docs/principles/actor-fidelity.md

@@ -125,7 +125,29 @@ a green run proves nothing.
 4. **Prompt guidance ships as a nudge whose take-rate is reported, never as the
    mechanism.** The evidence above is what that sentence is standing on.
 
-5. **Fail closed only on harness integrity, never on product semantics** — a lane that
+5. **Capability settings are recruiting decisions, not instrument settings.** Which
+   model runs a lane, and at what reasoning effort, changes who the participant IS —
+   the same way a persona prompt does. It is tempting to sort these into "the
+   participant" (persona) and "the harness" (model, effort) and treat a result that
+   moves with effort as a confound. That split does not survive contact: both are
+   things we set on a synthetic person before they touch the product, and the evidence
+   above says the persona half is the *weakly* grounded one. So a lane that abandons at
+   `medium` and completes at `high` produced two findings about two participants, not
+   one finding and one artifact.
+
+   What follows is a recording obligation, not a control obligation. Vary them
+   deliberately, declare what was recruited, and record the resolved value on the trace
+   (`ids.model`, `modelSettings.reasoningEffort`) so a reader can tell which participant
+   produced which result. The failure mode is not "we ran at medium" — it is running at
+   medium without saying so, then attributing the outcome to the one variable we did
+   declare.
+
+   **What this does not license:** claiming that any (model, effort) pair corresponds to
+   a real user population. No such mapping is calibrated here, and this project does not
+   assert one. A declared population says who was recruited; it does not say whom they
+   stand for.
+
+6. **Fail closed only on harness integrity, never on product semantics** — a lane that
    cannot report what it did is a broken instrument, which is a different thing from
    an actor that behaved unexpectedly.
 

package/docs/ramp/README.md

@@ -2,7 +2,7 @@
 
 Status: public-safe contributor and agent ramp.
 
-Package/source version in this tree: `0.55.0` (2026-08-20). The Observer is phone-usable as a stated requirement (observer/AGENTS.md); interactive primitives start from Base UI. The Observer renderer is the observer/ workspace artifact only; the legacy string-concat renderer was deleted at cutover (#426), and rollback is a version pin to 0.42.0. The containment boundary introduced in
+Package/source version in this tree: `0.56.0` (2026-08-21). The Observer is phone-usable as a stated requirement (observer/AGENTS.md); interactive primitives start from Base UI. The Observer renderer is the observer/ workspace artifact only; the legacy string-concat renderer was deleted at cutover (#426), and rollback is a version pin to 0.42.0. The containment boundary introduced in
 `0.15.1` remains in force: managed run and output paths bind to validated
 physical filesystem identities, and stored provider IDs are evidence, not
 cleanup authority. The bundled OSS meta-lab is dry-run only until

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "humanish",
-  "version": "0.55.0",
+  "version": "0.56.0",
   "description": "Open-source-safe CLI for persona simulation, observer review, and public-safe feedback drafts.",
   "author": "Daniel G Wilson <daniel@danielgwilson.com>",
   "keywords": [