@bendyline/gilde 0.1.17 → 0.1.18

package/authoring/gstack/evals/investigate.json

@@ -0,0 +1,153 @@
+{
+  "schemaVersion": 1,
+  "title": "Root-Cause Investigation — decimal cart total regression",
+  "objective": "Require the Root-Cause Investigation craftbook to reproduce a deterministic failure, identify the causal mechanism before editing, implement the smallest fix, and prove the regression test passes.",
+  "tags": [
+    "workflow",
+    "debugging",
+    "code-with-tests",
+    "node"
+  ],
+  "prompt": "Use the Root-Cause Investigation craftbook to diagnose the failing decimal cart-total behavior in this workspace. First reproduce it with the existing dependency-free test and preserve concrete evidence; do not edit until a causal hypothesis has been tested. Then make the smallest correct source change, keep the regression coverage, and write the craftbook's artifacts, including `reports/root-cause-investigation.md`. The final report must cite the files inspected, distinguish root cause from symptom, include the exact verification command and result, and describe a safe rollback.",
+  "setup": {
+    "projectName": "Decimal cart total regression",
+    "about": "A tiny dependency-free Node project with one reproducible arithmetic defect. The initial test is expected to fail.",
+    "missionObjectives": "Restore exact cent-based addition for decimal prices without broad refactoring and leave a passing regression test plus an evidence-backed investigation report.",
+    "files": [
+      {
+        "path": "package.json",
+        "content": "{\n  \"name\": \"cart-total-fixture\",\n  \"private\": true,\n  \"type\": \"module\"\n}\n"
+      },
+      {
+        "path": "src/cart-total.js",
+        "content": "/** Sum display-price strings and return a two-decimal total. */\nexport function cartTotal(prices) {\n  const cents = prices.reduce((sum, price) => sum + parseInt(price, 10) * 100, 0);\n  return (cents / 100).toFixed(2);\n}\n"
+      },
+      {
+        "path": "tests/cart-total.test.mjs",
+        "content": "import test from 'node:test';\nimport assert from 'node:assert/strict';\nimport { cartTotal } from '../src/cart-total.js';\n\ntest('keeps cents when totaling decimal display prices', () => {\n  assert.equal(cartTotal(['19.99', '5.50']), '25.49');\n});\n\ntest('handles a single whole-dollar price', () => {\n  assert.equal(cartTotal(['7.00']), '7.00');\n});\n"
+      },
+      {
+        "path": "source/incident.md",
+        "content": "# Incident note\n\nCheckout preview showed **$24.00** for line items **$19.99** and **$5.50**. The expected total is **$25.49**. The defect appeared after price inputs changed from integer-dollar strings to two-decimal display strings. No production data migration is in scope.\n"
+      },
+      {
+        "path": "tests/verify-cart-total.mjs",
+        "content": "import assert from 'node:assert/strict';\nimport { cartTotal } from '../src/cart-total.js';\n\nconst cases = [\n  [['19.99', '5.50'], '25.49'],\n  [['0.10', '0.20'], '0.30'],\n  [['12.34', '0.66'], '13.00'],\n  [['7.00'], '7.00'],\n  [['1.01', '2.02', '3.03'], '6.06'],\n];\nfor (const [prices, expected] of cases) assert.equal(cartTotal(prices), expected, prices.join(' + '));\nconsole.log('CART_TOTAL_ORACLE_OK 5/5');\n",
+        "surface": "harness"
+      }
+    ]
+  },
+  "mocks": [],
+  "success": {
+    "taskGraph": {
+      "requireCraftbookTask": true,
+      "requireTerminalStep": true
+    },
+    "summary": "The decimal regression is fixed by a diagnosed causal change, the existing test passes, and the craftbook report preserves reproducible evidence.",
+    "deliverables": [
+      {
+        "path": "reports/root-cause-investigation.md",
+        "kind": "markdown-report",
+        "minBytes": 1200,
+        "checks": [
+          {
+            "kind": "contains",
+            "file": "reports/root-cause-investigation.md",
+            "pattern": "^#{1,3}\\s+Root cause\\b[\\s\\S]*^#{1,3}\\s+(Fix|Changed files)\\b[\\s\\S]*^#{1,3}\\s+(Regression coverage|Verification)\\b[\\s\\S]*^#{1,3}\\s+Rollback\\b",
+            "flags": "im",
+            "label": "root cause through rollback sections"
+          },
+          {
+            "kind": "contains",
+            "file": "reports/root-cause-investigation.md",
+            "pattern": "parseInt|truncat(?:e|ed|ion)|discard(?:s|ed)?\\s+(?:the\\s+)?(?:decimal|fractional|cent)",
+            "flags": "i",
+            "label": "causal mechanism"
+          },
+          {
+            "kind": "contains",
+            "file": "reports/root-cause-investigation.md",
+            "pattern": "node\\s+--test\\s+tests/cart-total\\.test\\.mjs[\\s\\S]*(pass|2\\s+tests?|exit(?:ed)?\\s+0)",
+            "flags": "i",
+            "label": "verification command and result"
+          },
+          {
+            "kind": "citationsResolve",
+            "file": "reports/root-cause-investigation.md",
+            "minCitations": 2
+          }
+        ]
+      }
+    ],
+    "checks": [
+      {
+        "kind": "notContains",
+        "file": "src/cart-total.js",
+        "pattern": "\\bparseInt\\s*\\(",
+        "label": "lossy integer parsing removed"
+      },
+      {
+        "kind": "sourceParses",
+        "file": "src/cart-total.js"
+      },
+      {
+        "kind": "nodeRuns",
+        "file": "tests/cart-total.test.mjs",
+        "timeoutMs": 20000
+      },
+      {
+        "kind": "nodeScriptPasses",
+        "script": "tests/verify-cart-total.mjs",
+        "requiredOutput": [
+          {
+            "pattern": "CART_TOTAL_ORACLE_OK 5/5",
+            "label": "hidden multi-case regression oracle"
+          }
+        ]
+      }
+    ],
+    "unchangedFixtures": [
+      "package.json",
+      "tests/cart-total.test.mjs",
+      "source/incident.md"
+    ],
+    "taskNotes": {
+      "minBytes": 180,
+      "requireCraftbookTask": true,
+      "checks": [
+        {
+          "kind": "contains",
+          "file": "task-notes.md",
+          "pattern": "\\bDONE\\b[\\s\\S]*(node\\s+--test|tests/cart-total\\.test\\.mjs)[\\s\\S]*reports/root-cause-investigation\\.md",
+          "flags": "i",
+          "label": "terminal note records proof and report"
+        }
+      ]
+    }
+  },
+  "rubric": {
+    "artifact": {
+      "path": "reports/root-cause-investigation.md",
+      "kind": "markdown"
+    },
+    "axes": [
+      {
+        "name": "Causal rigor",
+        "description": "The report explains why the observed inputs fail in this implementation and distinguishes causal evidence from guesses."
+      },
+      {
+        "name": "Fix discipline",
+        "description": "The source change is minimal, directly addresses the diagnosed mechanism, and avoids unrelated refactoring."
+      },
+      {
+        "name": "Verification quality",
+        "description": "Reproduction, regression coverage, final command result, blast radius, and rollback are concrete and auditable."
+      }
+    ]
+  },
+  "qualityFocus": [
+    "Diagnosis precedes source edits",
+    "The fix preserves decimal cents and passes the dependency-free test",
+    "The final report cites real workspace files and includes rollback"
+  ]
+}

package/authoring/gstack/evals/office-hours.json

@@ -0,0 +1,141 @@
+{
+  "schemaVersion": 1,
+  "title": "Idea Office Hours — evidence-backed wedge decision",
+  "objective": "Exercise the Idea Office Hours craftbook end to end on a constrained product idea and require a grounded decision, explicit alternatives, and an auditable craftbook-task handoff.",
+  "tags": [
+    "workflow",
+    "product",
+    "ideation",
+    "decision"
+  ],
+  "prompt": "Use the Idea Office Hours craftbook to pressure-test the QuietQueue idea in `source/idea-context.md`. Do not jump straight to implementation. Work through the evidence, compare at least three materially different options (including no-build/manual), choose a narrow pilot wedge, and write the craftbook's decision artifacts. The final decision at `decisions/idea-office-hours.md` must cite the supplied source, state what remains assumed, name a concrete next real-world action with an owner and measurement, and avoid inventing customer evidence.",
+  "setup": {
+    "projectName": "QuietQueue idea review",
+    "about": "A hermetic product-decision exercise for a small clinic scheduling concept. Only the seeded evidence is authoritative.",
+    "missionObjectives": "Decide whether there is a narrow, testable pilot worth running without handling PHI or committing to a full scheduling product.",
+    "files": [
+      {
+        "path": "source/idea-context.md",
+        "content": "# QuietQueue idea context\n\n## Audience\nBoutique physical-therapy clinics with 3–12 practitioners. The immediate user is the front-desk coordinator; clinic owners approve spend.\n\n## Observed problem\nAcross four discovery calls, coordinators described reminder calling as repetitive and hard to finish during busy afternoons. A two-week manual tally recorded **31 no-shows across 420 appointments** and about **6 staff hours per week** spent on reminder calls. These are directional observations, not a statistically representative study.\n\n## Constraints\n- A pilot must launch within two weeks.\n- It must not ingest protected health information.\n- The clinic can export appointment first name, mobile number, and time into a temporary CSV only after its privacy lead approves the fields.\n- The team has one product engineer and no native-mobile capacity.\n- Budget for the pilot is $750.\n\n## Candidate ideas raised so far\n1. A full scheduling and messaging platform.\n2. A CSV-driven reminder assistant that drafts messages for staff approval.\n3. A better call-list spreadsheet and a scripted reminder process (no new product).\n\n## Unknowns\nWhether patients will opt into SMS, whether approval adds too much staff friction, and whether fewer no-shows would persist beyond the pilot.\n"
+      }
+    ]
+  },
+  "mocks": [],
+  "success": {
+    "taskGraph": {
+      "requireCraftbookTask": true,
+      "requireTerminalStep": true
+    },
+    "summary": "The named craftbook produces a grounded wedge decision and records its terminal task handoff.",
+    "deliverables": [
+      {
+        "path": "decisions/idea-office-hours.md",
+        "kind": "markdown-report",
+        "minBytes": 1100,
+        "checks": [
+          {
+            "kind": "contains",
+            "file": "decisions/idea-office-hours.md",
+            "pattern": "^#{1,3}\\s+(Problem|Problem framing)\\b[\\s\\S]*^#{1,3}\\s+(User evidence|Evidence)\\b[\\s\\S]*^#{1,3}\\s+(Chosen wedge|Decision)\\b",
+            "flags": "im",
+            "label": "problem, evidence, and chosen-wedge sections"
+          },
+          {
+            "kind": "contains",
+            "file": "decisions/idea-office-hours.md",
+            "pattern": "^#{1,3}\\s+Alternatives considered\\b[\\s\\S]*(manual|no[- ]build)[\\s\\S]*^#{1,3}\\s+Next real-world action\\b",
+            "flags": "im",
+            "label": "manual alternative and next action"
+          },
+          {
+            "kind": "valueGrounding",
+            "file": "decisions/idea-office-hours.md",
+            "facts": [
+              {
+                "id": "observed-no-shows",
+                "label": "observed no-show evidence",
+                "required": [
+                  "31\\s+no-shows?"
+                ]
+              },
+              {
+                "id": "observed-appointments",
+                "label": "observed appointment denominator",
+                "required": [
+                  "420\\s+appointments?"
+                ]
+              },
+              {
+                "id": "staff-time",
+                "label": "weekly reminder effort",
+                "required": [
+                  "6\\s+(staff\\s+)?hours?\\s+per\\s+week"
+                ]
+              },
+              {
+                "id": "pilot-duration",
+                "label": "two-week pilot constraint",
+                "required": [
+                  "two[- ]week|2[- ]week"
+                ]
+              },
+              {
+                "id": "pilot-data-boundary",
+                "label": "no-PHI pilot constraint",
+                "required": [
+                  "(no|without|not)\\s+(handling|ingesting|using)?\\s*PHI|protected health information"
+                ]
+              }
+            ]
+          },
+          {
+            "kind": "citationsResolve",
+            "file": "decisions/idea-office-hours.md",
+            "minCitations": 1
+          }
+        ]
+      }
+    ],
+    "taskNotes": {
+      "minBytes": 160,
+      "requireCraftbookTask": true,
+      "checks": [
+        {
+          "kind": "contains",
+          "file": "task-notes.md",
+          "pattern": "\\bDONE\\b[\\s\\S]*decisions/idea-office-hours\\.md",
+          "label": "terminal craftbook note names the decision"
+        }
+      ]
+    },
+    "unchangedFixtures": [
+      "source/idea-context.md"
+    ]
+  },
+  "rubric": {
+    "artifact": {
+      "path": "decisions/idea-office-hours.md",
+      "kind": "markdown"
+    },
+    "axes": [
+      {
+        "name": "Diagnostic depth",
+        "description": "The decision separates observed evidence from assumptions and identifies the most consequential unknowns before proposing a solution."
+      },
+      {
+        "name": "Option quality",
+        "description": "Alternatives are genuinely different, fairly compared, and include a credible manual or no-build path."
+      },
+      {
+        "name": "Decision usefulness",
+        "description": "The chosen wedge is narrow, constraint-aware, measurable, and leads to a concrete owner-backed next action."
+      }
+    ],
+    "contextNote": "Judge only against `source/idea-context.md`; unsupported market or customer claims are defects."
+  },
+  "qualityFocus": [
+    "Evidence is clearly distinguished from assumption",
+    "The selected wedge is smaller than a full scheduling platform",
+    "The next action has an owner, measurement, and decision-useful outcome"
+  ]
+}

package/authoring/gstack/evals/plan-ceo-review.json

@@ -0,0 +1,135 @@
+{
+  "schemaVersion": 1,
+  "title": "Executive-Level Review — analytics expansion decision",
+  "objective": "Test whether the Executive-Level Review craftbook challenges a proposal, compares lower-cost alternatives, and produces an evidence-anchored mode verdict with decision conditions.",
+  "tags": [
+    "workflow",
+    "strategy",
+    "executive-review",
+    "decision"
+  ],
+  "prompt": "Use the Executive-Level Review craftbook to review `source/analytics-proposal.md`. Challenge whether the proposed scope is the best way to reach the outcome; compare at least three alternatives including the two-day manual export and waiting. Produce all craftbook artifacts and a decision-ready `reviews/executive-level-review.md`. Use exactly one allowed mode verdict, score every required dimension from 0–10 with a factual basis, separate unknowns from evidence, and state explicit approve, pause, and stop conditions. Do not invent customer or financial evidence.",
+  "setup": {
+    "projectName": "Analytics expansion review",
+    "about": "A hermetic executive decision review. The seeded proposal is the complete evidence set.",
+    "missionObjectives": "Decide what, if anything, should be approved now while preserving the fastest credible route to learning.",
+    "files": [
+      {
+        "path": "source/analytics-proposal.md",
+        "content": "# Embedded analytics proposal\n\n## Requested decision\nApprove an eight-week build of customer-facing dashboards with scheduled PDF delivery and CSV export.\n\n## Evidence\nTwelve of 86 active customers requested better reporting in support conversations. Three customers agreed to a design pilot, but none signed a paid commitment. Support currently creates weekly exports for five customers.\n\n## Cost and constraints\n- Estimate: two engineers for 8 weeks.\n- New charting service: $4,000 per month at current volume.\n- Security review is complete; accessibility and data-retention reviews are not.\n- The enterprise renewal decision is in six weeks.\n- A manual scheduled-export workflow can be built in two engineer-days and would test delivery cadence, but not dashboard engagement.\n\n## Unknowns\nWillingness to pay, which metrics matter, PDF accessibility, and whether scheduled delivery or interactive dashboards drives retention.\n"
+      }
+    ]
+  },
+  "mocks": [],
+  "success": {
+    "taskGraph": {
+      "requireCraftbookTask": true,
+      "requireTerminalStep": true
+    },
+    "summary": "The named craftbook produces a grounded executive verdict, alternatives, scorecard, and terminal task record.",
+    "deliverables": [
+      {
+        "path": "reviews/executive-level-review.md",
+        "kind": "markdown-report",
+        "minBytes": 1400,
+        "checks": [
+          {
+            "kind": "contains",
+            "file": "reviews/executive-level-review.md",
+            "pattern": "^#{1,3}\\s+Executive verdict\\b[\\s\\S]*(SCOPE EXPANSION|SELECTIVE|HOLD|SCOPE REDUCTION)[\\s\\S]*^#{1,3}\\s+Scorecard\\b[\\s\\S]*^#{1,3}\\s+Alternatives\\b[\\s\\S]*^#{1,3}\\s+Recommendation\\b[\\s\\S]*^#{1,3}\\s+Decision conditions\\b",
+            "flags": "im",
+            "label": "verdict, scorecard, alternatives, recommendation, and conditions"
+          },
+          {
+            "kind": "contains",
+            "file": "reviews/executive-level-review.md",
+            "pattern": "(user value).*?(strategic fit).*?(evidence strength).*?(feasibility).*?(sequencing).*?(reversibility).*?(operating cost).*?(downside)",
+            "flags": "is",
+            "label": "complete executive scorecard"
+          },
+          {
+            "kind": "valueGrounding",
+            "file": "reviews/executive-level-review.md",
+            "facts": [
+              {
+                "id": "customer-requests",
+                "required": [
+                  "12\\s+(?:of\\s+86\\s+)?(?:active\\s+)?customers"
+                ]
+              },
+              {
+                "id": "pilot-evidence",
+                "required": [
+                  "3\\s+customers?[^.]{0,60}(pilot|agreed)"
+                ]
+              },
+              {
+                "id": "build-duration",
+                "required": [
+                  "8\\s+weeks?"
+                ]
+              },
+              {
+                "id": "operating-cost",
+                "required": [
+                  "\\$4,?000\\s+per\\s+month"
+                ]
+              },
+              {
+                "id": "cheap-alternative",
+                "required": [
+                  "(two|2)[ -]?(engineer[- ])?days?[\\s\\S]{0,120}(manual|scheduled)[ -]export"
+                ]
+              }
+            ]
+          },
+          {
+            "kind": "citationsResolve",
+            "file": "reviews/executive-level-review.md",
+            "minCitations": 1
+          }
+        ]
+      }
+    ],
+    "taskNotes": {
+      "minBytes": 160,
+      "requireCraftbookTask": true,
+      "checks": [
+        {
+          "kind": "contains",
+          "file": "task-notes.md",
+          "pattern": "\\bDONE\\b[\\s\\S]*reviews/executive-level-review\\.md",
+          "label": "terminal craftbook note names the review"
+        }
+      ]
+    },
+    "unchangedFixtures": [
+      "source/analytics-proposal.md"
+    ]
+  },
+  "rubric": {
+    "artifact": {
+      "path": "reviews/executive-level-review.md",
+      "kind": "markdown"
+    },
+    "axes": [
+      {
+        "name": "Premise challenge",
+        "description": "The review tests whether action is warranted now and compares materially different ways to learn or deliver value."
+      },
+      {
+        "name": "Evidence discipline",
+        "description": "Scores and conclusions trace to supplied facts, with unsupported assumptions and missing decisions called out."
+      },
+      {
+        "name": "Executive usefulness",
+        "description": "The verdict is decisive, sequenced, reversible where possible, and paired with clear decision-changing conditions."
+      }
+    ]
+  },
+  "qualityFocus": [
+    "One allowed mode verdict is selected",
+    "All scorecard dimensions have evidence-backed scores",
+    "The two-day alternative and missing willingness-to-pay evidence influence the decision"
+  ]
+}