@pmelab/gtd 41.0.0 → 42.0.0

package/src/workflows/feature/feature.test.ts

@@ -1,5 +1,5 @@
 import { describe, expect, it } from "vitest"
-import { runsItsScript, testRepo, type Actor } from "../../testkit/index.js"
+import { runsItsScript, testRepo } from "../../testkit/index.js"
 import { CONFIG_SCAFFOLD } from "../../install/index.js"
 import { reviewDoc } from "../shared/config.fixture.js"
 
@@ -75,10 +75,7 @@ describe("afterTail feedback lap", () => {
     `}`,
   ].join("\n")
 
-  const prompts: string[] = []
-
   const feedbackLap = async () => {
-    prompts.length = 0
     const repo = testRepo({
       workflow: TAIL,
       env: {
@@ -93,10 +90,7 @@ describe("afterTail feedback lap", () => {
     await repo.drive({
       "review-gate.check": () => undefined,
       "build.health.check": () => undefined,
-      "build.review.reviewing": (beat, r) => {
-        prompts.push(beat.content)
-        r.write(".gtd/REVIEW.md", review())
-      },
+      "build.review.composing": (_, r) => r.write(".gtd/REVIEW.md", review()),
     })
     repo.write(".gtd/REVIEW.md", review("also export multiply"))
     await repo.land()
@@ -137,35 +131,6 @@ describe("afterTail feedback lap", () => {
     expect(rest.content).not.toContain("write the change fresh")
     expect(repo.subjects().some((subject) => subject.includes("seam."))).toBe(false)
   })
-
-  it("the first review says everything since base, the one after feedback only what is new", async () => {
-    const repo = await feedbackLap()
-    expect(prompts[0]).toContain("it covers everything since the base")
-    expect(prompts[0]).not.toContain("only what is new since the previous review")
-    const writes: Record<string, () => [string, string]> = {
-      "design.triage": () => [
-        ".gtd/REQUIREMENTS.md",
-        "# Requirements\n\n- multiply, designed\n\n## Acceptance Criteria\n\n- AC-1: done\n",
-      ],
-      "implement.building": () => ["src/calc.ts", `${CALC}// again\n`],
-    }
-    const actors = new Proxy({} as Record<string, Actor>, {
-      get: (_, state: string): Actor | undefined =>
-        state === "build.review.await-review" || state === "idle"
-          ? undefined
-          : (beat, r) => {
-              if (state === "implement.coverage") return '[{"id":"AC-1","answer":"yes","p":0.9}]'
-              if (state === "build.review.reviewing") {
-                prompts.push(beat.content)
-                r.write(".gtd/REVIEW.md", review())
-              } else if (writes[state]) r.write(...writes[state]())
-              else if (state.endsWith("closing") || state === "re-unwind") r.runScript(beat.content)
-            },
-    })
-    await repo.drive(actors, { maxBeats: 80 })
-    expect(prompts[1]).toContain("it covers only what is new since the previous review")
-    expect(prompts[1]).not.toContain("everything since the base")
-  })
 })
 
 describe("an in-flight 28.0.0 naive-pass process", () => {

package/src/workflows/feature/feature.ts

@@ -9,7 +9,7 @@ import {
   slice,
   start,
 } from "../../flows/index.js"
-import { gate, healthy, type EscalationCount } from "../shared/health.js"
+import { gate, healthy } from "../shared/health.js"
 import { flagModeOn, shipped } from "../shared/options.js"
 import { FEEDBACK, fixCheck, QUALITY, REQUIREMENTS } from "../shared/steps.js"
 import { refusePendingHandOff, tail } from "../shared/tail.js"
@@ -135,8 +135,8 @@ const buildSlice = async (sliceBase: string, build: SliceBuild): Promise<void> =
   })
 }
 
-// The lap's findings the fix cap left must not reach the review, nor the next process's lap,
-// which appends to the file. Empty counts as resolved, as the lap reads it.
+// Between slices nothing reviews the lap's leftovers, and the next lap appends to the file.
+// Empty counts as resolved, as the lap reads it.
 const settleQuality = async (): Promise<void> => {
   let rested = false
   while ((read(QUALITY) ?? "").length > 0) {
@@ -151,15 +151,13 @@ const settleQuality = async (): Promise<void> => {
   if (rested) await healthy(() => fixCheck())
 }
 
-/** The quality lap, then the human review since `base`. */
-const reviewedTail = (base: string, lensBase: string, lap?: "feedback"): Promise<ReviewOutcome> =>
-  tail(
-    async (escalations: EscalationCount) => {
-      await settleQuality()
-      return review(base, escalations, { lap })
-    },
-    { green: true, lensBase },
-  )
+/** The triaged quality lap, then the human review since `base`. */
+const reviewedTail = (base: string, lensBase: string): Promise<ReviewOutcome> =>
+  tail((escalations, report) => review(base, escalations, { report }), {
+    green: true,
+    lensBase,
+    triage: true,
+  })
 
 // seam.naive / seam.judge never run on a feedback lap: its design → implement is built unjudged
 // and unrefactored, held to every criterion's coverage since the process start.
@@ -181,7 +179,7 @@ const featureFeedback = async (
     })
     const tier = riskTierOn() ? recordRisk() : undefined
     closed = await implementing({ tier, amend: true })
-    outcome = await reviewedTail(outcome.base, lensBase, "feedback")
+    outcome = await reviewedTail(outcome.base, lensBase)
   }
 }
 
@@ -201,7 +199,7 @@ export const afterTail = async (first: ReviewOutcome): Promise<void> => {
     await design(outcome.base)
     const tier = riskTierOn() ? recordRisk() : undefined
     await implement({ tier, amend: true })
-    outcome = await buildTail(outcome.base, { green: true, lap: "feedback" })
+    outcome = await buildTail(outcome.base, { green: true })
   }
 }
 

package/src/workflows/feature/flag-mode.feature

@@ -67,7 +67,7 @@ Feature: flag mode off lands the naive pass with no release flag, seam judge or
       """
     And step "build.quality.correctness.reviewing" changes nothing
     When the process is driven
-    Then the process rests at "build.review.reviewing"
+    Then the process rests at "build.review.composing"
     And "src/calc.ts" contains "subtract"
     And "src/calc.test.ts" contains "[AC-1] subtracts"
     And the commit subjects end with:
@@ -85,5 +85,5 @@ Feature: flag mode off lands the naive pass with no release flag, seam judge or
       gtd(check): implement.health.check → implement.coverage
       gtd(judge): implement.coverage → implement.closing
       gtd(check): implement.closing → build.quality.correctness.reviewing
-      gtd(agent): build.quality.correctness.reviewing → build.review.reviewing
+      gtd(agent): build.quality.correctness.reviewing → build.review.composing
       """

package/src/workflows/feature/lens-review.feature

@@ -0,0 +1,234 @@
+@inmem
+Feature: the quality lenses are the only agentic review
+
+  A judge classifies each `## ` finding the lenses wrote to `.gtd/QUALITY.md`:
+  a mechanical one is fixed by the fixer, noise is dropped with its `p` kept in
+  a trailer, and a design one becomes a note in `.gtd/REVIEW.md`. gtd writes
+  that file itself — no reviewer turn runs — and lists, without tick boxes,
+  what was fixed automatically and what was dropped as noise.
+
+  Background:
+    Given the shipped workflows, as gtd install scaffolds them
+    And an environment variable "GTD_QUALITYREVIEWS" set to "correctness"
+    And an environment variable "GTD_FLAGMODE" set to "false"
+    And a file "src/calc.ts" with:
+      """
+      export const add = (a: number, b: number) => a + b
+      """
+    And the working tree is committed as "feat: add calculator"
+    And a file "TODO.md" with:
+      """
+      # Add subtract
+
+      subtract(a, b) returns the difference.
+      """
+    And step "unwind" changes nothing
+    And step "start-gate.check" changes nothing
+    And step "design.triage" writes ".gtd/REQUIREMENTS.md" with:
+      """
+      # Requirements
+
+      ## Concerns
+
+      ### Subtract
+
+      - `subtract(a, b)` returns `a - b`.
+
+      ## Acceptance Criteria
+
+      - AC-1: subtract(a, b) returns a - b.
+      """
+    And step "design.gate.answer" changes nothing
+    And step "seam.naive" writes "src/calc.test.ts" with:
+      """
+      it("[AC-1] subtracts", () => {})
+      """
+    And step "seam.naive" writes "src/calc.ts" with:
+      """
+      export const add = (a: number, b: number) => a + b
+      export const subtract = (a: number, b: number) => a - b
+      """
+    And step "seam.check" changes nothing
+    And step "implement.applying" runs its script
+    And step "implement.health.check" changes nothing
+    And step "implement.closing" runs its script
+    And judge step "implement.coverage" answers:
+      """
+      [{"id":"AC-1","answer":"yes","p":0.9}]
+      """
+    And step "build.quality.correctness.reviewing" writes ".gtd/QUALITY.md" with:
+      """
+      # Quality
+
+      ## subtract has no negative test
+
+      Add a test subtracting a larger number.
+
+      ## add could be a method
+
+      A class would read better.
+
+      ## AC-1 ignores floating point
+
+      Should subtract round? The criterion does not say.
+      """
+    And step "build.quality.sorting" runs its script
+    And step "build.review.composing" runs its script
+
+  Scenario: a mechanical finding is fixed and listed, noise dropped and listed, a design one noted
+    Given judge step "build.quality.triage" answers:
+      """
+      [{"id":"Q1","answer":"mechanical","p":0.9},{"id":"Q2","answer":"noise","p":0.85},{"id":"Q3","answer":"design","p":0.8}]
+      """
+    And step "build.fix.quality.fixing" writes "src/calc.test.ts" with:
+      """
+      it("[AC-1] subtracts", () => {})
+      it("subtracts a larger number", () => {})
+      """
+    And step "build.fix.quality.fixing" deletes ".gtd/QUALITY.md"
+    And step "build.health.check" changes nothing
+    When the process is driven
+    Then the process rests at "build.review.await-review"
+    And ".gtd/QUALITY.md" does not exist
+    And the commit "gtd(judge): build.quality.triage → build.quality.sorting" has a body containing "Gtd-Mark: noise=Q2 p=0.85 add could be a method"
+    And ".gtd/REVIEW.md" contains:
+      """
+      ## AC-1 ignores floating point
+
+      Should subtract round? The criterion does not say.
+
+      ## Fixed automatically
+
+      - subtract has no negative test (
+      """
+    And ".gtd/REVIEW.md" contains the short hash of the commit "gtd(agent): build.fix.quality.fixing → build.health.check"
+    And ".gtd/REVIEW.md" contains:
+      """
+      ## Dropped as noise
+
+      - add could be a method (p=0.85)
+      """
+    And ".gtd/REVIEW.md" does not contain "[ ]"
+    And the commit subjects end with:
+      """
+      gtd(check): implement.closing → build.quality.correctness.reviewing
+      gtd(agent): build.quality.correctness.reviewing → build.quality.triage
+      gtd(judge): build.quality.triage → build.quality.sorting
+      gtd(check): build.quality.sorting → build.fix.quality.fixing
+      gtd(agent): build.fix.quality.fixing → build.health.check
+      gtd(check): build.health.check → build.review.composing
+      gtd(check): build.review.composing → build.review.await-review
+      """
+
+  Scenario: an unsure verdict makes a finding a note, never a fix or a drop
+    Given judge step "build.quality.triage" answers:
+      """
+      [{"id":"Q1","answer":"mechanical","p":0.5},{"id":"Q2","answer":"noise","p":0.6}]
+      """
+    When the process is driven
+    Then the process rests at "build.review.await-review"
+    And ".gtd/REVIEW.md" contains "## subtract has no negative test"
+    And ".gtd/REVIEW.md" contains "## add could be a method"
+    And ".gtd/REVIEW.md" contains "## AC-1 ignores floating point"
+    And ".gtd/REVIEW.md" does not contain "Fixed automatically"
+    And no commit body contains "Gtd-Mark: noise="
+    And the commit subjects end with:
+      """
+      gtd(judge): build.quality.triage → build.quality.sorting
+      gtd(check): build.quality.sorting → build.review.composing
+      gtd(check): build.review.composing → build.review.await-review
+      """
+
+  Scenario: a mechanical finding the fix cap leaves becomes a note
+    Given judge step "build.quality.triage" answers:
+      """
+      [{"id":"Q1","answer":"mechanical","p":0.9},{"id":"Q2","answer":"mechanical","p":0.9},{"id":"Q3","answer":"mechanical","p":0.9}]
+      """
+    And step "build.fix.quality.fixing" writes ".gtd/QUALITY.md" with:
+      """
+      # Quality
+
+      ## add could be a method
+
+      A class would read better.
+      """
+    And step "build.health.check" changes nothing
+    When the process is driven
+    Then the process rests at "build.review.await-review"
+    And ".gtd/QUALITY.md" does not exist
+    And ".gtd/REVIEW.md" contains:
+      """
+      ## add could be a method
+
+      _The fixer left this unresolved._
+
+      A class would read better.
+      """
+    And ".gtd/REVIEW.md" contains "- subtract has no negative test ("
+    And ".gtd/REVIEW.md" contains "- AC-1 ignores floating point ("
+    And the commit subjects end with:
+      """
+      gtd(judge): build.quality.triage → build.fix.quality.fixing
+      gtd(agent): build.fix.quality.fixing
+      gtd(agent): build.fix.quality.fixing
+      gtd(agent): build.fix.quality.fixing → build.health.check
+      gtd(check): build.health.check → build.review.composing
+      gtd(check): build.review.composing → build.review.await-review
+      """
+
+  Scenario: a fixer that renames a chunk it leaves counts nothing fixed
+    Given judge step "build.quality.triage" answers:
+      """
+      [{"id":"Q1","answer":"mechanical","p":0.9},{"id":"Q2","answer":"mechanical","p":0.9},{"id":"Q3","answer":"mechanical","p":0.9}]
+      """
+    And step "build.fix.quality.fixing" writes ".gtd/QUALITY.md" with:
+      """
+      # Quality
+
+      ## null check in parse()
+
+      Still open.
+      """
+    And step "build.health.check" changes nothing
+    When the process is driven
+    Then the process rests at "build.review.await-review"
+    And ".gtd/REVIEW.md" contains "## null check in parse()"
+    And ".gtd/REVIEW.md" contains "## subtract has no negative test"
+    And ".gtd/REVIEW.md" contains "check whether it was fixed"
+    And ".gtd/REVIEW.md" does not contain "Fixed automatically"
+
+  Scenario: prose the fixer leaves outside every chunk does not hide its fixes
+    Given judge step "build.quality.triage" answers:
+      """
+      [{"id":"Q1","answer":"mechanical","p":0.9},{"id":"Q2","answer":"mechanical","p":0.9},{"id":"Q3","answer":"mechanical","p":0.9}]
+      """
+    And step "build.fix.quality.fixing" writes ".gtd/QUALITY.md" with:
+      """
+      # Quality
+
+      All fixed.
+      """
+    And step "build.health.check" changes nothing
+    When the process is driven
+    Then the process rests at "build.review.await-review"
+    And ".gtd/REVIEW.md" contains "- subtract has no negative test ("
+    And ".gtd/REVIEW.md" does not contain "check whether it was fixed"
+
+  Scenario: signing off a composed review ends the process
+    Given judge step "build.quality.triage" answers:
+      """
+      [{"id":"Q1","answer":"noise","p":0.9},{"id":"Q2","answer":"noise","p":0.9},{"id":"Q3","answer":"noise","p":0.9}]
+      """
+    And step "build.review.closing" runs its script
+    When the process is driven
+    Then the process rests at "build.review.await-review"
+    When gtd lands
+    And the process is driven
+    Then the process rests at "idle"
+    And ".gtd/REVIEW.md" does not exist
+    And the commit subjects end with:
+      """
+      gtd(check): build.review.composing → build.review.await-review
+      gtd(human): build.review.await-review → build.review.closing
+      gtd(check): build.review.closing → idle
+      """

package/src/workflows/feature/one-step-land.feature

@@ -64,7 +64,7 @@ Feature: a naive pass the judge rules land is itself the implementation
 
   Scenario: a land judgment applies the naive pass and runs from idle to the review gate with no rebuild
     When the process is driven
-    Then the process rests at "build.review.reviewing"
+    Then the process rests at "build.review.composing"
     And "src/greeting.ts" contains "Hello, world"
     And "src/acceptance.test.ts" contains "[AC-1] holds"
     And the commit subjects end with:
@@ -81,7 +81,7 @@ Feature: a naive pass the judge rules land is itself the implementation
       gtd(check): implement.health.check → implement.coverage
       gtd(judge): implement.coverage → implement.closing
       gtd(check): implement.closing → build.quality.correctness.reviewing
-      gtd(agent): build.quality.correctness.reviewing → build.review.reviewing
+      gtd(agent): build.quality.correctness.reviewing → build.review.composing
       """
     And the commit "gtd(human): idle → unwind" has a body containing "Gtd-Var: flag_risk_tier=0"
     And the commit "gtd(agent): seam.naive → seam.check" has a body containing "Gtd-Kept:"

package/src/workflows/feature/quality-unresolved.feature

@@ -1,14 +1,16 @@
 @inmem
-Feature: quality findings left unresolved hold a feature process short of its review
+Feature: quality findings left unresolved between slices hold the next slice back
 
-  When the quality lap's fix cap leaves `.gtd/QUALITY.md` behind, the process
-  rests for a person instead of handing findings nobody acted on to the
-  human review.
+  Between slices no review reads the quality lap's leftovers, so when the fix
+  cap leaves `.gtd/QUALITY.md` behind the process rests for a person before
+  the next slice. The last slice's leftovers become review notes instead (see
+  `lens-review.feature`).
 
   Background:
     Given the shipped workflows, as gtd install scaffolds them
     And an environment variable "GTD_QUALITYREVIEWS" set to "correctness"
     And an environment variable "GTD_FLAGMODE" set to "false"
+    And an environment variable "GTD_SLICEMAXCRITERIA" set to "1"
     And a file "src/calc.ts" with:
       """
       export const add = (a: number, b: number) => a + b
@@ -16,9 +18,9 @@ Feature: quality findings left unresolved hold a feature process short of its re
     And the working tree is committed as "feat: add calculator"
     And a file "TODO.md" with:
       """
-      # Add subtract
+      # More arithmetic
 
-      subtract(a, b) returns the difference.
+      subtract and multiply.
       """
     And step "unwind" changes nothing
     And step "start-gate.check" changes nothing
@@ -26,41 +28,53 @@ Feature: quality findings left unresolved hold a feature process short of its re
       """
       # Requirements
 
-      ## Concerns
-
-      ### Subtract
-
-      - `subtract(a, b)` returns `a - b`.
-
       ## Acceptance Criteria
 
       - AC-1: subtract(a, b) returns a - b.
+      - AC-2: multiply(a, b) returns a * b.
       """
     And step "design.gate.answer" changes nothing
-    And step "seam.naive" writes "src/calc.test.ts" with:
+    And step "design.slicing" writes ".gtd/SLICES.md" with:
+      """
+      # Slices
+
+      ## Subtract
+
+      - AC-1
+
+      ## Multiply
+
+      - AC-2
+      """
+    And step "seam.naive", when prompted with "This slice builds AC-1 only", writes "src/sub.test.ts" with:
       """
       it("[AC-1] subtracts", () => {})
       """
-    And step "seam.naive" writes "src/calc.ts" with:
+    And step "seam.naive", when prompted with "This slice builds AC-2 only", writes "src/mul.test.ts" with:
       """
-      export const add = (a: number, b: number) => a + b
-      export const subtract = (a: number, b: number) => a - b
+      it("[AC-2] multiplies", () => {})
       """
     And step "seam.check" changes nothing
     And step "implement.applying" runs its script
     And step "implement.health.check" changes nothing
     And step "implement.closing" runs its script
-    And judge step "implement.coverage" answers:
+    And judge step "implement.coverage", when its questions contain "AC-1", answers:
       """
       [{"id":"AC-1","answer":"yes","p":0.9}]
       """
+    And judge step "implement.coverage", when its questions contain "AC-2", answers:
+      """
+      [{"id":"AC-2","answer":"yes","p":0.9}]
+      """
 
-  Scenario: findings the fix cap left unresolved rest at build.quality-unresolved
+  Scenario: findings the fix cap left after a slice rest at build.quality-unresolved
     Given step "build.quality.correctness.reviewing" writes ".gtd/QUALITY.md" with:
       """
       # Quality
 
-      - subtract has no test
+      ## subtract has no negative test
+
+      Add a test subtracting a larger number.
       """
     And step "build.fix.quality.fixing" from now on changes nothing
     And step "build.health.describe" writes ".gtd/ESCALATION.md" with:
@@ -74,33 +88,34 @@ Feature: quality findings left unresolved hold a feature process short of its re
     Given step "build.fix" writes "src/calc.ts" with:
       """
       export const add = (a: number, b: number) => a + b
-      export const subtract = (a: number, b: number) => a - b
       // checked
       """
     And step "build.health.check" changes nothing
     When gtd lands
     And the process is driven
     Then the process rests at "build.quality-unresolved"
-    And ".gtd/QUALITY.md" contains "subtract has no test"
+    And ".gtd/QUALITY.md" contains "subtract has no negative test"
 
     When gtd lands
     And the process is driven
     Then the process rests at "build.quality-unresolved"
 
-    Given a file "src/calc.ts" with:
-      """
-      export const add = (a: number, b: number) => a + b
-      export const subtract = (a: number, b: number) => a - b
-      // checked by hand
-      """
-    And the file ".gtd/QUALITY.md" is deleted
-    And step "build.health.check" changes nothing
+    Given the file ".gtd/QUALITY.md" is deleted
+    And step "build.quality.correctness.reviewing" from now on changes nothing
+    And step "build.review.composing" runs its script
     When gtd lands
     And the process is driven
-    Then the process rests at "build.review.reviewing"
+    Then the process rests at "build.review.await-review"
     And the commit subjects end with:
       """
-      gtd(check): build.health.check → build.quality-unresolved
       gtd(human): build.quality-unresolved → build.health.check
-      gtd(check): build.health.check → build.review.reviewing
+      gtd(check): build.health.check → seam.naive
+      gtd(agent): seam.naive → seam.check
+      gtd(check): seam.check → implement.applying
+      gtd(check): implement.applying → implement.health.check
+      gtd(check): implement.health.check → implement.coverage
+      gtd(judge): implement.coverage → implement.closing
+      gtd(check): implement.closing → build.quality.correctness.reviewing
+      gtd(agent): build.quality.correctness.reviewing → build.review.composing
+      gtd(check): build.review.composing → build.review.await-review
       """

package/src/workflows/feature/refactor-absorbs.feature

@@ -69,7 +69,7 @@ Feature: a refactor that carried the whole feature still gets a quality lap over
       """
     And step "refactor.health.check" changes nothing
     And step "refactor.closing" runs its script
-    And step "refactor.build.review.reviewing" writes ".gtd/REVIEW.md" with:
+    And step "refactor.build.review.composing" writes ".gtd/REVIEW.md" with:
       """
       # Review: abc1234
 
@@ -113,7 +113,7 @@ Feature: a refactor that carried the whole feature still gets a quality lap over
 
     Given step "build.quality.correctness.reviewing" changes nothing
     When the process is driven
-    Then the process rests at "build.review.reviewing"
+    Then the process rests at "build.review.composing"
     And "src/calc.ts" contains "subtract"
     And the commit subjects end with:
       """
@@ -123,7 +123,7 @@ Feature: a refactor that carried the whole feature still gets a quality lap over
       gtd(check): implement.health.check → implement.coverage
       gtd(judge): implement.coverage → implement.closing
       gtd(check): implement.closing → build.quality.correctness.reviewing
-      gtd(agent): build.quality.correctness.reviewing → build.review.reviewing
+      gtd(agent): build.quality.correctness.reviewing → build.review.composing
       """
     And the output lines of gtd "slices" start with:
       """

package/src/workflows/feature/refactor-first.feature

@@ -72,7 +72,7 @@ Feature: a refactor-first judgment walks one fixed path to the review gate
       """
     And step "refactor.health.check" changes nothing
     And step "refactor.closing" runs its script
-    And step "refactor.build.review.reviewing" writes ".gtd/REVIEW.md" with:
+    And step "refactor.build.review.composing" writes ".gtd/REVIEW.md" with:
       """
       # Review: abc1234
 
@@ -121,7 +121,7 @@ Feature: a refactor-first judgment walks one fixed path to the review gate
       [{"id":"AC-1","answer":"yes","p":0.9}]
       """
     When the process is driven
-    Then the process rests at "build.review.reviewing"
+    Then the process rests at "build.review.composing"
     And "src/calc.ts" contains "op(a, b, 1)"
     And "src/calc.ts" contains "subtract"
     And the commit subjects end with:
@@ -138,14 +138,14 @@ Feature: a refactor-first judgment walks one fixed path to the review gate
       gtd(check): refactor.deriving → refactor.building
       gtd(agent): refactor.building → refactor.health.check
       gtd(check): refactor.health.check → refactor.closing
-      gtd(check): refactor.closing → refactor.build.review.reviewing
-      gtd(agent): refactor.build.review.reviewing → refactor.build.review.await-review
+      gtd(check): refactor.closing → refactor.build.review.composing
+      gtd(check): refactor.build.review.composing → refactor.build.review.await-review
       gtd(human): refactor.build.review.await-review → refactor.build.review.closing
       gtd(check): refactor.build.review.closing → implement.building
       gtd(agent): implement.building → implement.health.check
       gtd(check): implement.health.check → implement.coverage
       gtd(judge): implement.coverage → implement.closing
-      gtd(check): implement.closing → build.review.reviewing
+      gtd(check): implement.closing → build.review.composing
       """
     And the commit "gtd(human): idle → unwind" has a body containing "Gtd-Var: flag_risk_tier=0"
     And the commit "gtd(agent): seam.naive → seam.check" has a body containing "Gtd-Kept:"

package/src/workflows/feature/refactor-rebuild.feature

@@ -83,7 +83,7 @@ Feature: refactor-first writes a refactor note, refactors on its own, then rebui
       """
     And step "refactor.health.check" changes nothing
     And step "refactor.closing" runs its script
-    And step "refactor.build.review.reviewing" writes ".gtd/REVIEW.md" with:
+    And step "refactor.build.review.composing" writes ".gtd/REVIEW.md" with:
       """
       # Review: abc1234
 
@@ -147,7 +147,7 @@ Feature: refactor-first writes a refactor note, refactors on its own, then rebui
       """
     And step "refactor.health.check" changes nothing
     And step "refactor.closing" runs its script
-    And step "refactor.build.review.reviewing" writes ".gtd/REVIEW.md" with:
+    And step "refactor.build.review.composing" writes ".gtd/REVIEW.md" with:
       """
       # Review: abc1234
 
@@ -193,7 +193,7 @@ Feature: refactor-first writes a refactor note, refactors on its own, then rebui
       """
     And step "refactor.health.check" changes nothing
     And step "refactor.closing" runs its script
-    And step "refactor.build.review.reviewing" writes ".gtd/REVIEW.md" with:
+    And step "refactor.build.review.composing" writes ".gtd/REVIEW.md" with:
       """
       # Review: abc1234
 
@@ -242,7 +242,7 @@ Feature: refactor-first writes a refactor note, refactors on its own, then rebui
       [{"id":"AC-1","answer":"yes","p":0.9}]
       """
     When the process is driven
-    Then the process rests at "build.review.reviewing"
+    Then the process rests at "build.review.composing"
     And "src/calc.ts" contains "op(a, b, -1)"
     And "src/acceptance.test.ts" contains "[AC-1]"
     And ".gtd/REFACTOR-NOTE.md" does not exist
@@ -253,14 +253,14 @@ Feature: refactor-first writes a refactor note, refactors on its own, then rebui
       gtd(check): refactor.deriving → refactor.building
       gtd(agent): refactor.building → refactor.health.check
       gtd(check): refactor.health.check → refactor.closing
-      gtd(check): refactor.closing → refactor.build.review.reviewing
-      gtd(agent): refactor.build.review.reviewing → refactor.build.review.await-review
+      gtd(check): refactor.closing → refactor.build.review.composing
+      gtd(check): refactor.build.review.composing → refactor.build.review.await-review
       gtd(human): refactor.build.review.await-review → refactor.build.review.closing
       gtd(check): refactor.build.review.closing → implement.building
       gtd(agent): implement.building → implement.health.check
       gtd(check): implement.health.check → implement.coverage
       gtd(judge): implement.coverage → implement.closing
-      gtd(check): implement.closing → build.review.reviewing
+      gtd(check): implement.closing → build.review.composing
       """
     And the output lines of gtd "slices" start with:
       """