Sync Yield from Intelligence Flow @ 4c338abc0d14

README.md

@@ -81,8 +81,8 @@ use the documentation by job:
 - [tutorials](docs/tutorials/README.md) — review, approval, environment
   repair, bounded debugging, and migration;
 - [examples](docs/examples.md) — working programs in all four languages;
-- [evaluations](evals/README.md) — pinned conversion cases, reproducible
-  measurements, and the raw-evidence publication boundary;
+- [evaluations](evals/README.md) — first-party workflow conformance and runtime
+  invariant results, including the exact claim boundary;
 - [convert an existing skill](docs/convert-existing-skill.md) — move
   control flow into code without claiming that fixture execution proves
   every reading of the original prose;

UPSTREAM.json

@@ -2,11 +2,11 @@
   "files": {
     ".gitignore": "803c5f79d6da7f2c5a1dc0ce27c53b8e5c059309b165782831c4c472af058a4c",
     "LICENSE": "fff261ce507eabd57666c283a621f33e183a3aedebda04c4ecbc6309a62f5edf",
-    "README.md": "d3a13ebcd016ab1ba6a4fac64ca4a555ff738e9ce4c5c9744ee4e046e9dc08ef",
+    "README.md": "acd9f37f0f7013a4bf99df382fd3b5c14d070914ec958e3198d9c70aa8317ce6",
     "cmd/yskill/main.go": "a88bdd133118aa7b86cdf021e67052bd8e64e63b3792c091f7612699af89648b",
     "docs/README.md": "cd8a4105f4f04143172661b39789e3c533d031ad46437aa2c078625a29a0a9db",
     "docs/convert-existing-skill.md": "0dd538e908a1f2d3a1958e4b2cc1efede2a8e7a73341e32c1a75fec69a9acb8a",
-    "docs/examples.md": "8c81f990b8d3a06b42cedcf34813a11dbe0606201782f72903a8c14d91d1ae23",
+    "docs/examples.md": "3c19bacae7ec7b31bf93417e61d01cd1872f26228fedf0ed6decfce7c41ed274",
     "docs/locus-conformance.md": "a71fd5a72678dcad344afc71cd7a788090ef008a8e5e82d925da5a3db41af6a0",
     "docs/locus-converter.md": "b8d1bdb63cb0283ace524f572fb13f0d1cb535b130c5ea3e112d2efae55a8681",
     "docs/locus-yield.md": "c29d0c5801f32591828fc348385fb25c1850b39afc22770098eae0614fc3a417",
@@ -40,33 +40,13 @@
     "docs/tutorials/data-migration.md": "6699fc47cda84c4f46a3703ca4a69bcced7cc4a9df339a308e807ea3a99578bd",
     "docs/tutorials/environment-repair.md": "289c9b261e3768c2c58d60d9ef7a09e85f89038438a08febb4b9c12ebada88d1",
     "evals/.gitignore": "0d5020173666118bafe31c857b96aa325809f41d159ca51324ceaf239e043347",
-    "evals/README.md": "c6fb08381877d3cddce22afe59746902ba73deebd78e52a414f85bcbbf441b22",
-    "evals/cases/README.md": "03bd5a5bca02c9052a8413b5d30f463d37e30d130a99093d9642a1e0bbede29d",
-    "evals/cases/actions-auditor/README.md": "e9efdfa7657840ed1829b2831abc1ef988c7f3faac5de98e9444d5751df9b913",
-    "evals/cases/actions-auditor/SKILL.md": "bdc81816f25144bcc44badb14ff3b46a234365466a8eeb6b3a6f6128ed57fd9b",
-    "evals/cases/actions-auditor/workflow.ts": "622f6dd037548a5ae0eb097a77392dbcc92eaa294691a4c4e30c4abfaa9544ec",
-    "evals/cases/doc-coauthoring/README.md": "d72fc22a0d7134c0764efb190fad7f7a6a7d93b0f55ce09447f469cb1711ecda",
-    "evals/cases/doc-coauthoring/SKILL.md": "8ab3e342ef202f22fe2763d1d5e0d5d6918ccdf5a30426abfc02f71b8a0b00d8",
-    "evals/cases/doc-coauthoring/workflow.ts": "df78a605f5c0d7e52ba63ef2b91757dda1f4a5229fbe66fcb7c9adcb9aac9efb",
-    "evals/cases/gstack-review/README.md": "fbfcba7aca73c0fc13813e3e85edf86b4988ac3e0e538209d8d3a89df7e35cd2",
-    "evals/cases/gstack-review/SKILL.md": "9b4914d4e569cf70fdc794d88c1c1714d4e88b4646ab9762d29da2cc2b892fba",
-    "evals/cases/gstack-review/workflow.ts": "0743492bc14aeacd941cfbe90ef909ffe47f474941fd1cc28c33a38147ae6910",
-    "evals/cases/index.json": "b363426b292939f977360daaaad3ea4875a447f59657fed74275f6cc4084de43",
-    "evals/cases/mcp-builder/README.md": "db708ccf258d6888173e573c02455a40c689ab51e09f7c57311fe1cefea32f14",
-    "evals/cases/mcp-builder/SKILL.md": "0a5c45ae7ad2e902f2a5c2c782e78db995b8e42869f36fad42839993e346dd5e",
-    "evals/cases/mcp-builder/workflow.ts": "fdf1ed76dc0cbe0eafd95ce1c8da7f8e65419e1f655c2df9ce88eaf285eee8a5",
-    "evals/cases/systematic-debugging/README.md": "5654154dd172c6405fcd8833d73c2ec2deb273416208b29890eb35b84fa0b438",
-    "evals/cases/systematic-debugging/SKILL.md": "2d130b11d680d43f02ef126a77b462c585101e4df1632b89873597acdd7b189b",
-    "evals/cases/systematic-debugging/workflow.ts": "86aaacc7d1fab4e320e5ce0fab7bd9972c137d6b17e8ebddc6d118299d2a1a80",
-    "evals/cases/vercel-deploy/README.md": "e585f09b81b3a3a2f23624f1d8e7c816af5b50d88f4204513072128bdef39b0c",
-    "evals/cases/vercel-deploy/SKILL.md": "4ac80ad71f42c327f63685530b6b98ce3e287459527d2c0bafb5e1c716c313da",
-    "evals/cases/vercel-deploy/workflow.ts": "78667cea9b4b4d3dfc54ef6944ca9aa75d837b0170a4d3811b8f55172f63d6f2",
-    "evals/package-lock.json": "22c099aa5f2d9959084d6703e4a94ff7fa34216cac21b9d81137e2dd155c09d5",
-    "evals/package.json": "3361b8d579664bf9d6642eaeca6a906a1bca43fd0b1b543b00b2d5e0aa021b17",
-    "evals/results/README.md": "8eda7b901660466e60e5e32609cb2a3dcca7d9d512a93f2f379090b16da15525",
-    "evals/results/latest.json": "339a15841921fa270c668aa5c55cd33a271493735b4b85beaeaba1b25807216a",
-    "evals/scripts/measure-source.mjs": "064a62073422c067e12b0f2b77cc65a38f7d3440efb1a91020595a941470db7a",
-    "evals/scripts/validate.mjs": "2616fdf2ebbfd940dff7e06bcb33f4a362e063484a4beb62f54ef2856747abd7",
+    "evals/README.md": "7b0a92908d4eff84f1716756a00536ed79b31ef542928a2fd0c17c234161bba6",
+    "evals/package-lock.json": "cfbd68d590e92b94233a807fd774e6667e9474096ab76e1c5cd037acfb4e3200",
+    "evals/package.json": "aa97f75cadbf6d1802dcf03f4b34dcbbc4aea8c2b5ef18130af26a9e63ecac51",
+    "evals/results/README.md": "65b74bfab83dfc51fc5b8a3a4824c37b722fb3a358a5dcfb217dc2af199f4801",
+    "evals/results/latest.json": "cc9346c0db8e8395464a7348ea440a891fad0080350ac5d677b1a2875047e4d4",
+    "evals/scripts/run.mjs": "119248cd4d383326d396f4459df28c51ce7c31088d97639851f2f5ab4c329892",
+    "evals/scripts/validate.mjs": "7ac2a3239cd1b52a604c8a8a52a1af4ac55c4007ac2751d4fbfe20d0e1f568dc",
     "examples/convert-skill/SKILL.md": "e6376f34365d4ac030d316db55e91f0c606a501668099f4ecf1e27d43ec806a2",
     "examples/convert-skill/fixtures/responses.json": "5b5f27b0ae5962360ac7b0e779992f430c655f352a38da37debb3505c32cd60a",
     "examples/convert-skill/main.go": "eb987ed1ce9139e6d0d40a2fae20c532c2fba3b27357e31555db2f0f52db4e39",
@@ -286,6 +266,7 @@
     "release-notes/2026-08-01-evaluation-case-guides.md": "570bc996eb4d2892456a02d938fb6299d106c431bb842ce726db1df550f779d7",
     "release-notes/2026-08-01-evaluation-surface.md": "64fb5e2fdad3ccd41967e028cf4675c45939174000281c443f4b28d96dc07549",
     "release-notes/2026-08-01-example-library.md": "14d6ca40529a6aeeb72872295e57bb0d7dcc7824df31e029e497602060ce4c97",
+    "release-notes/2026-08-01-first-party-evaluations.md": "9e74c48112343c409a88358138db80d731004f87900f1b23073f16c415163fcf",
     "release-notes/2026-08-01-initial-projection.md": "d38f0832b5552fb97237b30d19bb63369442ddde69e678b752ac07eedeb7ba3d",
     "release-notes/2026-08-01-multi-language-and-converter.md": "d0cf62d191e6a58e827f3b35d442f88b83be4dde98aa25ad54e4ae3ff787a453",
     "release-notes/2026-08-01-remove-stray-analysis-traces.md": "0567f78ee97ffd23b3f26b5c39606e9ff6659c50a3fdef04ed9b3aa86cfa99af",
@@ -302,7 +283,7 @@
   "generator": "operatorstack/yield:project",
   "schema_version": 1,
   "source": {
-    "commit": "381aea9c6d153503e68ad96e82ac236a354db432",
+    "commit": "4c338abc0d14f017eb3501885adca7594bd5b6f4",
     "path": "labs/22-yield",
     "repository": "operatorstack/intelligence-flow"
   }

docs/examples.md

@@ -61,14 +61,12 @@ When adapting an example, change the repository-specific commands and model
 instructions. Keep stable operation IDs for existing steps so saved runs can
 replay them.
 
-## Conversion cases used by evaluations
+## Evaluated examples
 
-The [evaluation cases](../evals/cases/) are different from the example library.
-Each one pins a public third-party skill to an exact commit and digest, then
-shows the smaller model-facing `SKILL.md` beside the TypeScript workflow that
-owns its order, commands, gates, and completion.
+The [first-party evaluation suite](../evals/) runs every example-library
+pattern through TypeScript, Python, Go, and Rust. It also checks resume, replay,
+changed behavior, blocking, and changed source.
 
-The source-size harness and early summaries live in [`evals/`](../evals/).
-Large transcripts and temporary repositories are external artifacts, never
-committed source. A behavioral result is publishable only when its summary
-binds the exact artifact URI and SHA-256.
+These results show that the tested Yield version runs the workflow steps in
+code. They do not compare Yield with prose or test whether an agent's judgment
+is correct.

evals/README.md

@@ -1,39 +1,55 @@
 # Yield evaluations
 
-This directory contains the public, reviewable part of Yield's evaluation
-system: case definitions, pinned source identities, conversion programs,
-measurement code, validation rules, and small result summaries.
+These evaluations test Yield itself. They do not compare Yield with another
+tool, company, prompt, or skill.
 
-Raw agent transcripts, temporary repositories, command logs, and model
-responses do not belong in Git. A full campaign uploads those files as one
-immutable artifact bundle and records its URI and SHA-256 in the published
-summary. Until that bundle exists, the summary must say `unpublished`.
+The suite answers two questions:
 
-## Layout
+1. Can each checked-in example workflow reach its expected final result
+   through every supported SDK?
+2. Does the runtime behave correctly when a run resumes, replays, blocks, or
+   encounters changed code?
 
-- `cases/` — pinned public source identities plus the thin skill and Yield
-  program used for each conversion.
-- `scripts/measure-source.mjs` — reproduces the source-size comparison from
-  pinned upstream files.
-- `scripts/validate.mjs` — fail-closed validation for cases and summaries.
-- `results/latest.json` — small website-safe summary. It is not raw evidence.
-- `runs/` — local or CI output; ignored by Git and projected releases.
+## Current coverage
 
-## Run
+- 10 workflow patterns written by this project.
+- 4 SDKs: TypeScript, Python, Go, and Rust.
+- 40 end-to-end workflow tests.
+- 5 runtime checks: resume and complete, repeat the same saved step, stop when
+  behavior changes, block when a rule fails, and require approval for changed
+  source.
+
+Run the exact suite and refresh the checked-in result:
+
+```bash
+cd evals
+npm run eval
+```
+
+Check that the published result still matches the current source:
 
 ```bash
-npm install
-npm run validate
-npm run measure
+npm test
 ```
 
-`npm run measure` writes a fresh summary under `runs/`. Publishing that result
-requires a separate promotion step that binds the raw artifact digest, the
-exact Yield commit, model identity, harness version, and case-set digest.
+## What a passing result proves
+
+A passing result proves that the tested Yield revision:
+
+- executes each owned workflow test to `completed`;
+- runs command steps rather than asking the model to invent their outputs;
+- presents requests in the program-defined order;
+- resumes from recorded responses;
+- returns to the same saved step during replay;
+- stops on changed behavior or failed requirements.
+
+## What it does not prove
 
-## Claim boundary
+This suite does not prove that Yield is better than prose, that an agent's
+judgment is correct, or that illustrative commands are production-safe. The
+Fixed test data supplies agent and human responses so the suite can test only
+the code-controlled workflow layer.
 
-Source-size measurements show how much model-facing text and workflow source
-the prototypes contain. They do not prove behavioral equivalence. Behavioral
-claims require executable fixtures, held-out oracles, repeated model runs, and
-the immutable raw artifact named by the result summary.
+`results/latest.json` is a compact, website-safe result. Its source hash is
+computed from the CLI, engine, protocol, SDKs, example workflows, fixtures, and
+evaluation harness. CI reruns the suite instead of trusting that file alone.