feat(goals): gate acceptance vectors for mappable scenario directives + nightly dream-cycle evaluate step (age-26y1)

.github/workflows/nightly.yml

@@ -251,6 +251,35 @@ jobs:
           fi
           scripts/nightly-knowledge-cycle.sh record-stage dream-cycle ok "summary at dream-cycle/summary.md"
 
+      - name: Dream cycle — scenario satisfaction (report-only)
+        if: steps.precondition.outputs.decision == 'RUN'
+        env:
+          NIGHTLY_KNOWLEDGE_CYCLE_DIR: ${{ runner.temp }}/knowledge-cycle
+        run: |
+          # Producer side of the satisfaction panel (ag-26y1): evaluate every
+          # directive-linked scenario against its acceptance vectors and append
+          # the verdicts to .agents/rpi/scenario-results.json so satisfaction is
+          # measured nightly alongside the dream cycle. Runs at the repo root —
+          # it needs GOALS.md, spec/scenarios/, and the real gate scripts — NOT
+          # inside nightly-dream-cycle.sh, which snapshots .agents/ into an
+          # ephemeral, GOALS-less workspace and is deliberately non-mutating.
+          # REPORT-ONLY: a gate-shaped scenario that fails its own threshold must
+          # not fail the cycle — mirror the compile-mine warn-on-nonzero guard.
+          SCEN_DIR="${RUNNER_TEMP}/knowledge-cycle/scenario-satisfaction"
+          mkdir -p "$SCEN_DIR"
+          set +e
+          ./cli/bin/ao eval scenario evaluate --all --json > "$SCEN_DIR/report.json"
+          rc=$?
+          if [[ $rc -ne 0 ]]; then
+            scripts/nightly-knowledge-cycle.sh record-stage scenario-satisfaction warn "ao eval scenario evaluate exited $rc (report-only; does not block the cycle)"
+          else
+            written=$(jq -r '.written // 0' "$SCEN_DIR/report.json" 2>/dev/null || echo 0)
+            gate_pass=$(jq -r '[.evaluations[] | select(.shape == "gate" and .verdict == "pass")] | length' "$SCEN_DIR/report.json" 2>/dev/null || echo 0)
+            gate_total=$(jq -r '[.evaluations[] | select(.shape == "gate")] | length' "$SCEN_DIR/report.json" 2>/dev/null || echo 0)
+            scripts/nightly-knowledge-cycle.sh record-stage scenario-satisfaction ok "wrote ${written} result(s); ${gate_pass}/${gate_total} gate-shaped scenarios satisfied"
+          fi
+          exit 0
+
       - name: Athena follow-up (placeholder — shares compile substrate)
         if: steps.precondition.outputs.decision == 'RUN'
         env:
@@ -269,6 +298,7 @@ jobs:
         run: |
           scripts/nightly-knowledge-cycle.sh record-stage compile skip "${{ steps.precondition.outputs.reason }}"
           scripts/nightly-knowledge-cycle.sh record-stage dream-cycle skip "${{ steps.precondition.outputs.reason }}"
+          scripts/nightly-knowledge-cycle.sh record-stage scenario-satisfaction skip "${{ steps.precondition.outputs.reason }}"
           scripts/nightly-knowledge-cycle.sh record-stage athena skip "${{ steps.precondition.outputs.reason }}"
 
       - name: Append knowledge-cycle summary to step summary

spec/scenarios/s-2026-05-24-002.json

@@ -7,6 +7,9 @@
   "narrative": "A user or evaluator exercises the system behavior for this goal: Given the three install scripts (install.sh, install-codex.sh, install-opencode.sh), when the install-smoke CI gate runs tests/install/test-install-smoke.sh, then all three scripts pass syntax+structure validation and, when cli/bin/ao exists, ao --version / ao help / and the flywheel, goals, inject subcommands are confirmed registered..",
   "expected_outcome": "The implementation satisfies the goal in observable behavior: Given the three install scripts (install.sh, install-codex.sh, install-opencode.sh), when the install-smoke CI gate runs tests/install/test-install-smoke.sh, then all three scripts pass syntax+structure validation and, when cli/bin/ao exists, ao --version / ao help / and the flywheel, goals, inject subcommands are confirmed registered..",
   "satisfaction_threshold": 0.9,
+  "acceptance_vectors": [
+    { "dimension": "correctness", "threshold": 1.0, "check": "gate:install-smoke" }
+  ],
   "source": "human",
   "status": "active"
 }

spec/scenarios/s-2026-05-24-003.json

@@ -7,6 +7,9 @@
   "narrative": "A user or evaluator exercises the system behavior for this goal: Given the tests/_quarantine/ directory, when the validate-quarantine-empty CI gate runs, then it finds zero active quarantined test suites (the directory stays empty or every entry is promoted, deleted, or tracked as explicit follow-up)..",
   "expected_outcome": "The implementation satisfies the goal in observable behavior: Given the tests/_quarantine/ directory, when the validate-quarantine-empty CI gate runs, then it finds zero active quarantined test suites (the directory stays empty or every entry is promoted, deleted, or tracked as explicit follow-up)..",
   "satisfaction_threshold": 1,
+  "acceptance_vectors": [
+    { "dimension": "correctness", "threshold": 1.0, "check": "gate:quarantine-empty" }
+  ],
   "source": "human",
   "status": "active"
 }

spec/scenarios/s-2026-05-24-004.json

@@ -7,6 +7,9 @@
   "narrative": "A user or evaluator exercises the system behavior for this goal: Given a single learning in the knowledge corpus, when scripts/check-flywheel-lifecycle.sh runs, then all five stages (capture, retrieval, inject, round-trip, citation) are traced and the flywheel-lifecycle CI gate passes..",
   "expected_outcome": "The implementation satisfies the goal in observable behavior: Given a single learning in the knowledge corpus, when scripts/check-flywheel-lifecycle.sh runs, then all five stages (capture, retrieval, inject, round-trip, citation) are traced and the flywheel-lifecycle CI gate passes..",
   "satisfaction_threshold": 0.9,
+  "acceptance_vectors": [
+    { "dimension": "correctness", "threshold": 1.0, "check": "gate:flywheel-lifecycle" }
+  ],
   "source": "human",
   "status": "active"
 }

spec/scenarios/s-2026-05-24-005.json

@@ -7,6 +7,9 @@
   "narrative": "A user or evaluator exercises the system behavior for this goal: Given the cli/ and cli/internal/ Go packages, when the go-complexity CI gate runs golangci-lint, then zero functions exceed cyclomatic complexity 20 (cli/) and 18 (cli/internal/)..",
   "expected_outcome": "The implementation satisfies the goal in observable behavior: Given the cli/ and cli/internal/ Go packages, when the go-complexity CI gate runs golangci-lint, then zero functions exceed cyclomatic complexity 20 (cli/) and 18 (cli/internal/)..",
   "satisfaction_threshold": 1,
+  "acceptance_vectors": [
+    { "dimension": "maintainability", "threshold": 1.0, "check": "gate:go-complexity-ceiling" }
+  ],
   "source": "human",
   "status": "active"
 }

spec/scenarios/s-2026-05-24-007.json

@@ -7,6 +7,9 @@
   "narrative": "A user or evaluator exercises the system behavior for this goal: Given a change that introduces codex parity drift, when the codex-parity CI gate (scripts/audit-codex-parity.sh) runs at push time, then it reports zero parity findings and blocks the push if any drift exists..",
   "expected_outcome": "The implementation satisfies the goal in observable behavior: Given a change that introduces codex parity drift, when the codex-parity CI gate (scripts/audit-codex-parity.sh) runs at push time, then it reports zero parity findings and blocks the push if any drift exists..",
   "satisfaction_threshold": 1,
+  "acceptance_vectors": [
+    { "dimension": "consistency", "threshold": 1.0, "check": "gate:codex-parity-drift" }
+  ],
   "source": "human",
   "status": "active"
 }

spec/scenarios/s-2026-05-24-010.json

@@ -7,6 +7,9 @@
   "narrative": "A user or evaluator exercises the system behavior for this goal: Given the workbench-behavioral-v1 eval suite (12 tasks), when make -C evals/workbench verify runs, then golden solutions score 12/12 pass and broken variants are detected 12/12, and the eval-skill-delta CI gate validates the structure..",
   "expected_outcome": "The implementation satisfies the goal in observable behavior: Given the workbench-behavioral-v1 eval suite (12 tasks), when make -C evals/workbench verify runs, then golden solutions score 12/12 pass and broken variants are detected 12/12, and the eval-skill-delta CI gate validates the structure..",
   "satisfaction_threshold": 0.9,
+  "acceptance_vectors": [
+    { "dimension": "correctness", "threshold": 1.0, "check": "gate:eval-workbench-verify" }
+  ],
   "source": "human",
   "status": "active"
 }

spec/scenarios/s-2026-05-24-012.json

@@ -7,6 +7,9 @@
   "narrative": "A user or evaluator exercises the system behavior for this goal: Given a fresh checkout after a runtime-state wipe, when the .agents/ snapshot/restore mechanism (soc-rv5p) rehydrates from durable storage, then scripts/corpus-stats.sh reports non-degraded learning/pattern/decision counts and the corpus-freshness gate confirms coverage..",
   "expected_outcome": "The implementation satisfies the goal in observable behavior: Given a fresh checkout after a runtime-state wipe, when the .agents/ snapshot/restore mechanism (soc-rv5p) rehydrates from durable storage, then scripts/corpus-stats.sh reports non-degraded learning/pattern/decision counts and the corpus-freshness gate confirms coverage..",
   "satisfaction_threshold": 0.85,
+  "acceptance_vectors": [
+    { "dimension": "durability", "threshold": 1.0, "check": "gate:corpus-freshness" }
+  ],
   "source": "human",
   "status": "active"
 }

spec/scenarios/s-2026-05-24-015.json

@@ -7,6 +7,9 @@
   "narrative": "A user or evaluator exercises the system behavior for this goal: Given a surface the teardown (epic ag-097) explicitly removed and recorded in scripts/removed-apparatus.txt, when an /evolve cycle or any change reintroduces that surface, then the no-apparatus-regrowth gate fails — teardown-removed apparatus stays removed..",
   "expected_outcome": "The implementation satisfies the goal in observable behavior: Given a surface the teardown (epic ag-097) explicitly removed and recorded in scripts/removed-apparatus.txt, when an /evolve cycle or any change reintroduces that surface, then the no-apparatus-regrowth gate fails — teardown-removed apparatus stays removed..",
   "satisfaction_threshold": 0.85,
+  "acceptance_vectors": [
+    { "dimension": "regression", "threshold": 1.0, "check": "gate:no-apparatus-regrowth" }
+  ],
   "source": "human",
   "status": "active"
 }