Nightly 2026-04-20 — 2 cycles, +15 goal weight, 1 dream finding closed

cli/embedded/skills/standards/references/test-pyramid.md

@@ -57,7 +57,7 @@ The Traditional Pyramid          The AI-Native Shape
 │  regression guards, L0 for contracts, L3 for subsystems.  │
 │                                                           │
 │  L0: Contract — from SPEC.md alone                        │
-│  L1: Unit     — always write for regression safety         │
+│  L1: Unit     — always write for regression safety        │
 │  L2: Integration — DEFAULT for all agent-written tests    │
 │  L3: Component — agent writes, human defines scenarios    │
 ├───────────────────────────────────────────────────────────┤

cli/internal/provenance/provenance.go

@@ -45,12 +45,50 @@ type AuditReport struct {
 // fields the audit cares about. Additional fields in the source are
 // ignored.
 type learningFrontmatter struct {
-	Title      string `yaml:"title"`
-	SourceBead string `yaml:"source_bead"`
-	Source     string `yaml:"source"`
-	Date       string `yaml:"date"`
+	Title      string         `yaml:"title"`
+	SourceBead string         `yaml:"source_bead"`
+	Source     flexibleString `yaml:"source"`
+	Date       string         `yaml:"date"`
 }
 
+// flexibleString holds a learning frontmatter value that may be either a
+// plain YAML scalar or a structured map/sequence. Older learnings used
+// a bare string such as `source: retro-quick`; newer ones occasionally
+// use a map like `source: {session: ..., evidence: [...]}` to carry
+// provenance context. Either shape should be honored by the audit; the
+// canonical "missing source" check only asks whether the value is empty.
+type flexibleString string
+
+// UnmarshalYAML accepts scalar, map, or sequence nodes. Scalars decode
+// to their literal value. Maps and sequences decode to a non-empty
+// sentinel so the caller's MissingSources check does not mark the
+// field as empty; the exact serialization is not interpreted further.
+func (f *flexibleString) UnmarshalYAML(node *yaml.Node) error {
+	if node == nil {
+		*f = ""
+		return nil
+	}
+	switch node.Kind {
+	case yaml.ScalarNode:
+		*f = flexibleString(node.Value)
+		return nil
+	case yaml.MappingNode, yaml.SequenceNode:
+		*f = "<structured>"
+		return nil
+	case yaml.AliasNode:
+		if node.Alias != nil {
+			return f.UnmarshalYAML(node.Alias)
+		}
+		*f = ""
+		return nil
+	default:
+		*f = ""
+		return nil
+	}
+}
+
+func (f flexibleString) String() string { return string(f) }
+
 // Audit scans .agents/learnings/ under cwd and returns an
 // AuditReport. Never prints to stdout/stderr. Soft-fails on
 // individual file read/parse errors; returns a hard error only
@@ -102,7 +140,7 @@ func Audit(cwd string) (*AuditReport, error) {
 		}
 
 		// Missing source: empty source_bead AND empty source fields.
-		if strings.TrimSpace(fm.SourceBead) == "" && strings.TrimSpace(fm.Source) == "" {
+		if strings.TrimSpace(fm.SourceBead) == "" && strings.TrimSpace(fm.Source.String()) == "" {
 			report.MissingSources++
 		}
 

cli/internal/provenance/provenance_test.go

@@ -4,6 +4,7 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"strings"
 	"testing"
 	"time"
 )
@@ -513,3 +514,45 @@ func TestProvenanceAudit_NoAgentsDir(t *testing.T) {
 			report.StaleCitations, report.MissingSources)
 	}
 }
+
+// Regression: a learning whose frontmatter uses a structured (map or
+// sequence) `source:` value must still parse and must not be flagged as
+// a missing source. Pre-fix the YAML decoder returned
+// "cannot unmarshal !!map into string" and the file was skipped.
+func TestProvenanceAudit_StructuredSourceAccepted(t *testing.T) {
+	cwd := t.TempDir()
+	learnings := filepath.Join(cwd, ".agents", "learnings")
+	if err := os.MkdirAll(learnings, 0o755); err != nil {
+		t.Fatal(err)
+	}
+
+	recent := time.Now().AddDate(0, 0, -1).Format("2006-01-02")
+	mapSource := "---\ntitle: Structured\ndate: " + recent + "\nsource:\n  session: 2026-01-01 demo\n  evidence:\n    - a.md\n    - b.md\n---\n\nbody\n"
+	seqSource := "---\ntitle: Sequence\ndate: " + recent + "\nsource:\n  - a.md\n  - b.md\n---\n\nbody\n"
+	stringSource := "---\ntitle: Scalar\ndate: " + recent + "\nsource: retro-quick\n---\n\nbody\n"
+
+	mustWrite := func(name, body string) {
+		if err := os.WriteFile(filepath.Join(learnings, name), []byte(body), 0o644); err != nil {
+			t.Fatal(err)
+		}
+	}
+	mustWrite("map-source.md", mapSource)
+	mustWrite("seq-source.md", seqSource)
+	mustWrite("scalar-source.md", stringSource)
+
+	report, err := Audit(cwd)
+	if err != nil {
+		t.Fatalf("Audit: %v", err)
+	}
+	if report == nil {
+		t.Fatal("nil report")
+	}
+	for _, note := range report.Degraded {
+		if strings.Contains(note, "yaml:") {
+			t.Errorf("unexpected YAML degraded entry: %q", note)
+		}
+	}
+	if report.MissingSources != 0 {
+		t.Errorf("expected 0 missing sources for structured+scalar sources, got %d", report.MissingSources)
+	}
+}

skills-codex/.agentops-manifest.json

@@ -692,7 +692,7 @@
       "name": "crank",
       "source_skill": "skills/crank",
       "source_hash": "928d1301b7f3bf12607e779cdec279924b5b29d681faffd01e6819651def92e7",
-      "generated_hash": "63340a4214c757133c74ede362ed165cb55e32e1ca1909f0329d2155dc6e5695"
+      "generated_hash": "2a0078618d49f2048978451334d01c363202836165802eb0f13268dec5f9ffa5"
     },
     {
       "name": "deps",
@@ -704,7 +704,7 @@
       "name": "design",
       "source_skill": "skills/design",
       "source_hash": "",
-      "generated_hash": "0623a63763fc0737ecb29686b689531574966933160d1fb37f963606e7e5f66f"
+      "generated_hash": "f3089a1425339cf222d714f2451216ba683474765f56bed4c644cbf76f168b92"
     },
     {
       "name": "discovery",
@@ -776,7 +776,7 @@
       "name": "implement",
       "source_skill": "skills/implement",
       "source_hash": "2ab197d1810fd3eb56f73aac4645c02edc01aa41e524064b12b993c6861ebc32",
-      "generated_hash": "69fa694eaad00ead6eebce2590e981eca07660fba16887eb28f4e833e93eaf21"
+      "generated_hash": "aa7f88ad80c6fb371f277c90d2a9e45985f481cb5319e9e87ef137ea4aabea9f"
     },
     {
       "name": "inject",
@@ -884,7 +884,7 @@
       "name": "quickstart",
       "source_skill": "skills/quickstart",
       "source_hash": "68a6463ab7dded893cbdd27d7697c70d871ed4149b5c7cbe8bf20369c1e0a813",
-      "generated_hash": "92759426fabae70bba9045ff448df760e9a7470b8395310018a8e78957949b20"
+      "generated_hash": "3069919895447f39b60d3f19a37bd96ec4d5b2f869e6266f2ad07c15870241b6"
     },
     {
       "name": "ratchet",
@@ -908,7 +908,7 @@
       "name": "red-team",
       "source_skill": "skills/red-team",
       "source_hash": "e7ae62c4a820cf3b7c93e1e0e5913e046013d9dfe8ff48ffa9f85294376bb037",
-      "generated_hash": "0b59c335567b529586535737af1d8ab5b70e3d73c39bf9b1da1ac54cf73796a5"
+      "generated_hash": "345e71a7c9d165615b9f10d4c1826e1b3203e36e5aee03cdaa64ea682556bb3a"
     },
     {
       "name": "refactor",
@@ -986,7 +986,7 @@
       "name": "standards",
       "source_skill": "skills/standards",
       "source_hash": "837a4c5f624c5de7dbe37193fb1671a97d2f4c9381355c4b3745e247b632577a",
-      "generated_hash": "33afeca1a2dc5ba35f1fec148563937f4730d83f17b414cf71e2c40cfab5b68b"
+      "generated_hash": "a0eeea87ffddd155c82fc10eee522d3d36cfd054067b03af72da292c08cd3ed3"
     },
     {
       "name": "status",
@@ -1028,7 +1028,7 @@
       "name": "validation",
       "source_skill": "skills/validation",
       "source_hash": "2e71f8dbe94c65971f0dff7f82ac83396649ff036bbea388060174945f660c5c",
-      "generated_hash": "5cd2d80a2406d05f0eff3a7276ac14d46d00e66d8e0d5785c784bcaa0a8de463"
+      "generated_hash": "4c1bf07ba20d26e8c82d40272dcfbd520241f9033454c4bf2730a59defaf8d40"
     },
     {
       "name": "vibe",

skills-codex/crank/.agentops-generated.json

@@ -3,5 +3,5 @@
   "source_skill": "skills/crank",
   "layout": "modular",
   "source_hash": "928d1301b7f3bf12607e779cdec279924b5b29d681faffd01e6819651def92e7",
-  "generated_hash": "63340a4214c757133c74ede362ed165cb55e32e1ca1909f0329d2155dc6e5695"
+  "generated_hash": "2a0078618d49f2048978451334d01c363202836165802eb0f13268dec5f9ffa5"
 }

skills-codex/design/.agentops-generated.json

@@ -3,5 +3,5 @@
   "source_skill": "skills/design",
   "layout": "modular",
   "source_hash": "",
-  "generated_hash": "0623a63763fc0737ecb29686b689531574966933160d1fb37f963606e7e5f66f"
+  "generated_hash": "f3089a1425339cf222d714f2451216ba683474765f56bed4c644cbf76f168b92"
 }

skills-codex/design/SKILL.md

@@ -63,7 +63,7 @@ Compute the average score across all five dimensions.
 Invoke council with the product preset. See [references/product-council-preset.md](references/product-council-preset.md) for judge configuration.
 
 ```
-Skill(skill="council", args="--preset=product validate design alignment for: <goal>")
+$council --preset=product validate design alignment for: <goal>
 ```
 
 Pass the alignment matrix from Step 2 as context to the council judges.

skills-codex/implement/.agentops-generated.json

@@ -3,5 +3,5 @@
   "source_skill": "skills/implement",
   "layout": "modular",
   "source_hash": "2ab197d1810fd3eb56f73aac4645c02edc01aa41e524064b12b993c6861ebc32",
-  "generated_hash": "69fa694eaad00ead6eebce2590e981eca07660fba16887eb28f4e833e93eaf21"
+  "generated_hash": "aa7f88ad80c6fb371f277c90d2a9e45985f481cb5319e9e87ef137ea4aabea9f"
 }

skills-codex/implement/SKILL.md

@@ -199,7 +199,7 @@ Or use `$test <feature>` to auto-generate test candidates, then hand-refine.
 If skip conditions above are NOT met AND `--no-lifecycle` is NOT set:
 
 ```
-Skill(skill="test", args="generate <feature-scope> --quick")
+$test generate <feature-scope> --quick
 ```
 
 The generated test request must preserve the selected `test_levels` and BF expectations from Step 3.6. Review the generated tests. Adjust as needed (tests are MUTABLE in this context). If `$test` fails to produce useful output or is unavailable, fall back to manual test writing in Step 3.6 above.

skills-codex/quickstart/.agentops-generated.json

@@ -3,5 +3,5 @@
   "source_skill": "skills/quickstart",
   "layout": "modular",
   "source_hash": "68a6463ab7dded893cbdd27d7697c70d871ed4149b5c7cbe8bf20369c1e0a813",
-  "generated_hash": "92759426fabae70bba9045ff448df760e9a7470b8395310018a8e78957949b20"
+  "generated_hash": "3069919895447f39b60d3f19a37bd96ec4d5b2f869e6266f2ad07c15870241b6"
 }

skills-codex/red-team/.agentops-generated.json

@@ -3,5 +3,5 @@
   "source_skill": "skills/red-team",
   "layout": "modular",
   "source_hash": "e7ae62c4a820cf3b7c93e1e0e5913e046013d9dfe8ff48ffa9f85294376bb037",
-  "generated_hash": "0b59c335567b529586535737af1d8ab5b70e3d73c39bf9b1da1ac54cf73796a5"
+  "generated_hash": "345e71a7c9d165615b9f10d4c1826e1b3203e36e5aee03cdaa64ea682556bb3a"
 }

skills-codex/red-team/SKILL.md

@@ -194,7 +194,7 @@ When the same finding appears from multiple personas:
 Run council with red-team preset to review and consolidate all findings:
 
 ```
-Skill(skill="council", args="--preset=red-team [--quick] validate .agents/red-team/")
+$council --preset=red-team [--quick] validate .agents/red-team/
 ```
 
 Use `--quick` by default. Use full council (omit `--quick`) when `--deep` flag is set.

skills-codex/standards/.agentops-generated.json

@@ -3,5 +3,5 @@
   "source_skill": "skills/standards",
   "layout": "modular",
   "source_hash": "837a4c5f624c5de7dbe37193fb1671a97d2f4c9381355c4b3745e247b632577a",
-  "generated_hash": "33afeca1a2dc5ba35f1fec148563937f4730d83f17b414cf71e2c40cfab5b68b"
+  "generated_hash": "a0eeea87ffddd155c82fc10eee522d3d36cfd054067b03af72da292c08cd3ed3"
 }

skills-codex/validation/.agentops-generated.json

@@ -3,5 +3,5 @@
   "source_skill": "skills/validation",
   "layout": "modular",
   "source_hash": "2e71f8dbe94c65971f0dff7f82ac83396649ff036bbea388060174945f660c5c",
-  "generated_hash": "5cd2d80a2406d05f0eff3a7276ac14d46d00e66d8e0d5785c784bcaa0a8de463"
+  "generated_hash": "4c1bf07ba20d26e8c82d40272dcfbd520241f9033454c4bf2730a59defaf8d40"
 }

skills-codex/validation/SKILL.md

@@ -50,7 +50,7 @@ Skip silently if ao is unavailable or returns no results.
 **Run every step in order. Do not stop between steps.**
 
 ```
-STEP 1  ──  Skill(skill="vibe", args="recent [--quick]")
+STEP 1  ──  $vibe recent [--quick]
               Use --quick for fast/standard. Full council for full.
               PASS/WARN? → continue
               FAIL?      → write summary, output <promise>FAIL</promise>, stop
@@ -75,20 +75,20 @@ STEP 1.7 ── Lifecycle Checks (advisory except critical dependency findings)
               On budget expiry: skip remaining sub-steps, write [TIME-BOXED].
 
               a) if lifecycle tier >= minimal AND test_framework_detected:
-                   Skill(skill="test", args="coverage --quick")
+                   $test coverage --quick
                    Append coverage delta to phase summary.
 
               b) if lifecycle tier >= standard AND dependency_manifest_exists:
-                   Skill(skill="deps", args="vuln --quick")
+                   $deps vuln --quick
                    CRITICAL vulns (CVSS >= 9.0): **FAIL** (block shipping). Opt-out: `--allow-critical-deps` for acknowledged risk acceptance.
                    Non-critical: advisory note only.
 
               c) if lifecycle tier >= standard:
-                   Skill(skill="review", args="--diff --quick")
+                   $review --diff --quick
                    Append review findings to summary as advisory.
 
               d) if lifecycle tier == full AND modified_files_touch_hot_path:
-                   Skill(skill="perf", args="profile --quick")
+                   $perf profile --quick
                    Append perf findings to summary as advisory.
                    Hot path detection: modified files match benchmark files
                    or patterns (handler, middleware, router, parser, engine,
@@ -123,15 +123,15 @@ STEP 1.8 ── Stage 4: Behavioral Validation (holdout scenarios + agent-built
             FAIL? → write summary, output <promise>FAIL</promise>, stop
 
 STEP 2  ──  if epic_id:
-              Skill(skill="post-mortem", args="<epic-id> [--quick]")
+              $post-mortem <epic-id> [--quick]
             else:
-              Skill(skill="post-mortem", args="recent [--quick]")
+              $post-mortem recent [--quick]
               Use --quick for fast/standard. Full council for full.
               PASS/WARN? → continue
               FAIL?      → write summary, output <promise>FAIL</promise>, stop
 
 STEP 3  ──  if not --no-retro:
-              Skill(skill="retro")
+              $retro
 
 STEP 4  ──  if not --no-forge AND ao available:
               if [ -n "${CODEX_THREAD_ID:-}" ] || [ "${CODEX_INTERNAL_ORIGINATOR_OVERRIDE:-}" = "Codex Desktop" ]; then