fix(campaign): keep profile matrix evidence exact

.evolve/skill-runs.jsonl

@@ -49,3 +49,5 @@
 {"skill":"/agent-eval","ts":"2026-07-30T11:43:12Z","project":"agent-eval-profile-axis","target":"expandProfileAxes canonical harness propagation","operatorPrompt":"","durationMin":null,"verdict":"PASS","dispatchedTo":"/verify","operatorOverride":null,"transcriptPath":null,"traceDir":null}
 {"skill":"/verify","ts":"2026-07-30T11:43:12Z","project":"agent-eval-profile-axis","target":"agent-eval expandProfileAxes harness source fix","operatorPrompt":"","durationMin":null,"verdict":"PASS","dispatchedTo":"/stop","operatorOverride":null,"transcriptPath":null,"traceDir":null}
 {"skill":"/verify","ts":"2026-08-03T22:53:14Z","project":"agent-eval-official-optimizer-observations-20260803","target":"Agent Eval optimizer execution-owner contract","operatorPrompt":"","durationMin":null,"verdict":"passed","dispatchedTo":"/ship","operatorOverride":null,"transcriptPath":null,"traceDir":null}
+{"skill":"/agent-eval","ts":"2026-08-04T04:23:29Z","project":"agent-eval-profile-matrix-hardening-20260804","target":"profile matrix model evidence, cancellation, and cache identity","operatorPrompt":"","durationMin":null,"verdict":"PASS","dispatchedTo":"/verify","operatorOverride":null,"transcriptPath":null,"traceDir":null}
+{"skill":"/verify","ts":"2026-08-04T04:23:29Z","project":"agent-eval-profile-matrix-hardening-20260804","target":"agent-eval profile matrix hardening 0.144.3","operatorPrompt":"","durationMin":null,"verdict":"PASS","dispatchedTo":"/ship","operatorOverride":null,"transcriptPath":null,"traceDir":null}

CHANGELOG.md

@@ -4,6 +4,19 @@ All notable changes to `@tangle-network/agent-eval` and its sibling `agent-eval-
 
 ---
 
+## [0.144.3] - 2026-08-03 - exact profile matrix evidence
+
+### Changed
+
+- Campaign cells expose every distinct agent receipt model in `resolvedModels` and expose `resolvedModel` only when all agent receipts agree.
+- Snapshot validation accepts Router `-MMDD` model snapshots while rejecting routing selectors such as `@preset/name`.
+
+### Fixed
+
+- `runProfileMatrix` rejects mismatched receipt models, multiple resolved snapshots within one profile, and duplicate profile identities before they can corrupt comparisons.
+- A failed profile campaign now cancels active sibling campaigns through their existing abort signals instead of allowing additional paid work to continue.
+- Profile campaign cache identity now includes the caller commit, optional `dispatchRef`, profile identity, and comparison config, so changed execution cannot reuse and relabel an old cell.
+
 ## [0.144.2] - 2026-08-03 - concurrent exact profile comparison
 
 ### Changed

clients/python/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "agent-eval-rpc"
-version = "0.144.2"
+version = "0.144.3"
 description = "Python RPC client, official optimizer bridge, and DSPy metric adapter for @tangle-network/agent-eval."
 readme = "README.md"
 requires-python = ">=3.10"

clients/python/src/agent_eval_rpc/__init__.py

@@ -53,7 +53,7 @@
 try:
     __version__ = version("agent-eval-rpc")
 except PackageNotFoundError:
-    __version__ = "0.144.2"
+    __version__ = "0.144.3"
 
 __all__ = [
     "Client",

docs/eval-surface-map.md

@@ -19,7 +19,9 @@ composition point already exists (see Produced-state grading below).**
 When variants of the same task run inside one `runCampaign`, give those scenarios the same `seedGroup` so each repetition uses common randomness.
 Use `runProfileMatrix` instead when profiles are separate campaign axes.
 Set `maxConcurrency` for cases within one profile and `maxProfileConcurrency` for independent profile columns; results retain caller order regardless of completion order.
-An execution profile may carry the provider-facing moving model alias only when each paid-call receipt supplies the related snapshot-bearing model stored in the resulting `RunRecord`.
+Every paid-call receipt must match the profile, and a moving alias must resolve to one snapshot across the entire profile column.
+The caller commit and profile config are always part of cache identity; set `dispatchRef` when execution behavior can change without a new commit.
+A failed profile cancels active sibling columns before the matrix rejects.
 | `runAgentMatrix` | The bare N-axis cartesian scheduler with concurrency control. The layer beneath the eval surface: reach for it only when you need raw scheduling, not eval semantics. | cell results |
 
 Mental model: **measure** (`runCampaign`/`runEval`) → **factor** (`runProfileMatrix`) →

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tangle-network/agent-eval",
-  "version": "0.144.2",
+  "version": "0.144.3",
   "description": "Evaluate and improve AI agents from runs, traces, judges, and feedback. Compare candidates, cluster failures, measure lift, and gate releases.",
   "homepage": "https://github.com/tangle-network/agent-eval#readme",
   "repository": {

src/analyst/benchmark-implementation.ts

@@ -10,7 +10,7 @@ export const ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES = Object.freeze([
 ])
 
 export const ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256 =
-  'bc15c7ea5cc3e78c081a8a8ec94e5588a344a77d0410fc6746bdce5de886b4cf'
+  'f0d788f72ea83b6bd485f2ae61e58dda7e1d87ec79284ce245b3a7f0cc7ca812'
 
 /** The published benchmark evidence was produced at this package version, by
  * the retired one-shot direct runner, before trace analysts moved to the
@@ -117,7 +117,7 @@ export const ANALYST_BENCHMARK_IMPLEMENTATION_FILES = Object.freeze([
 ])
 
 export const ANALYST_BENCHMARK_IMPLEMENTATION_SHA256 =
-  'd317d9e7593ddc13f22a4a1e3d62d2fa1e4591395ef3984a0c7585a75794cf8d'
+  '67cd83aaaa158d44a9fa32e172cbe68fa97f01ae673ccb117469688b628bb9b1'
 
 export function analystBenchmarkImplementationDigest() {
   return ANALYST_BENCHMARK_IMPLEMENTATION_SHA256