@tekyzinc/gsd-t 4.0.12 → 4.0.14

package/CHANGELOG.md

@@ -2,6 +2,31 @@
 
 All notable changes to GSD-T are documented here. Updated with each release.
 
+## [4.0.14] - 2026-05-29 (M67 Scan Deep Document Phase — patch)
+
+### Added — deterministic deep document cross-population in the scan Workflow
+
+M66 made the tech-debt register deep but left living-document cross-population as a non-deterministic "lead-agent follow-on" — effectively dropped, a regression vs the old prose scan's Step 5. M67 adds a `Document` phase to `templates/workflows/gsd-t-scan.workflow.js` (between Synthesis and Render) that fans out **one agent per document**, each drawing on the same slices + verified findings the finders produced, so the docs are as thorough as the register.
+
+- `templates/workflows/gsd-t-scan.workflow.js`: new `Document` phase produces `docs/{architecture,workflows,infrastructure,requirements}.md` + `README.md` (merge-not-overwrite, Edit-not-Write on existing files) and the five `.gsd-t/scan/{architecture,security,quality,business-rules,contract-drift}.md` dimension files in the renderer's parsed formats. Component map + a user journey per feature-domain slice. Synthesis no longer writes the dimension files (moved to the doc stage); runs before Render so the HTML report reads the deep `architecture.md`.
+- Red Team (FAIL → GRUDGING-PASS, 1 fix cycle): HIGH-1 doc-clobber data-loss → deterministic `.gsd-t/scan/.doc-backup/` snapshot before the fan-out; HIGH-2 render double-count → grand total written as a `| Grand Total | N files | LOC |` table row that the parser short-circuits on (verified 1809, not 1919); LOW → dropped `-f` from git add + pathspec-excluded and gitignored `.doc-backup`.
+- `commands/gsd-t-scan.md`, `commands/gsd-t-help.md`, `templates/CLAUDE-global.md`: doc-ripple for the new phase. `.gitignore`: `.gsd-t/scan/.doc-backup/`.
+
+Suite: 1267 pass / 0 fail / 4 skip — zero regressions.
+
+## [4.0.13] - 2026-05-29 (M66 Scan Volume-Scaled Workflow Migration — patch)
+
+### Fixed — scan-depth regression: scan now fans out by codebase volume, not a fixed 5-teammate count
+
+Scan was the only major phase never migrated to a native Workflow. It hard-coded exactly 5 teammates (one per dimension) with zero volume scaling — a 5-file repo and a 1,809-file repo both got 5 agents, so a single `quality` agent sampled the top ~5 issues across the whole codebase and stopped. On a large codebase that produced a cursory 16-item register where a deep per-domain scan surfaced 117 findings. It also referenced a retired `autoSpawnHeadless()` + `headless-default-contract v2.0.0`.
+
+- `templates/workflows/gsd-t-scan.workflow.js` (NEW): `preflight → volume-probe (derives a per-area slice list, scaling 1-3 slices tiny → 15-40 large) → pipeline(per-slice deep finder "enumerate don't sample" → single verify) → deterministic JS archive + TD-numbering → synthesis (dedup/merge/re-rank) → render`. Slice depth is budget-aware.
+- `commands/gsd-t-scan.md`: rewritten as a thin `Workflow({scriptPath, args})` invoker; dead `autoSpawnHeadless`/`headless-default-contract v2.0.0` references stripped.
+- Red Team (FAIL → GRUDGING-PASS, 2 fix cycles): prose-driven archive/TD-numbering → deterministic JS (`fs.renameSync` collision-safe + `_parseMaxTd` + prior register content threaded into synthesis); render hollow-report guard + prompt↔parser format alignment; pipeline stage-2 defensive signature; probe model haiku→sonnet.
+- Doc-ripple: removed retired fixed-5-teammate scan-dimension references from `gsd-t-{complete-milestone,gap-analysis,feature,init-scan-setup}.md`; `templates/CLAUDE-global.md` + `commands/gsd-t-help.md`.
+
+Suite: 1267 pass / 0 fail / 4 skip — zero regressions.
+
 ## [4.0.12] - 2026-05-29
 
 ### Removed — M61-carryover test cleanup (green the suite)

package/commands/gsd-t-complete-milestone.md

@@ -463,19 +463,13 @@ node scripts/gsd-t-watch-state.js advance --agent-id "$GSD_T_AGENT_ID" --parent-
 
 Before tagging, ensure scan docs reflect the final state of the codebase after all milestone work. This is the full-refresh counterpart to the micro-updates done during execute/quick/debug.
 
-If `.gsd-t/scan/` exists (a prior scan has been run):
-1. Check `.gsd-t/scan/.cache.json` for staleness — count commits since last scan
-2. If **any dimension is stale** (>0 commits since scan):
-   - Log: "Running milestone scan checkpoint — refreshing all stale scan dimensions..."
-   - Re-run all stale dimensions using scan teammates (same team structure as `gsd-t-scan` Step 2):
-     - Stale `architecture.md` → architecture teammate (model: haiku)
-     - Stale `quality.md` → quality teammate (model: sonnet)
-     - Stale `security.md` → security teammate (model: sonnet)
-     - Stale `business-rules.md` → business-rules teammate (model: haiku)
-     - Stale `contract-drift.md` → contracts teammate (model: haiku)
-   - Update `.gsd-t/scan/.cache.json` after refresh
+If `.gsd-t/techdebt.md` exists (a prior scan has been run):
+1. Check staleness — count commits since the register was last written (use `.gsd-t/scan/.cache.json` if present, else the register's git mtime).
+2. If **stale** (>0 commits since scan):
+   - Log: "Running milestone scan checkpoint — refreshing the tech-debt register..."
+   - Re-run the scan by invoking the volume-scaled scan Workflow (`templates/workflows/gsd-t-scan.workflow.js`, same as `/gsd-t-scan`) — the probe re-slices the codebase and the finders refresh the register. This replaces the retired fixed-5-teammate dimension refresh.
    - Update `.gsd-t/techdebt.md` — mark any items resolved during this milestone as `[RESOLVED]`
-3. If all dimensions are fresh → skip with log: "Scan docs are fresh — no checkpoint refresh needed"
+3. If fresh → skip with log: "Scan register is fresh — no checkpoint refresh needed"
 
 If `.gsd-t/scan/` doesn't exist → skip (no scan data to maintain).
 

package/commands/gsd-t-feature.md

@@ -14,12 +14,8 @@ If `.gsd-t/scan/.cache.json` exists:
 1. Read the cache and check `scannedAt` for each dimension
 2. Count commits since the scan: `git rev-list --count --after="{scannedAt}" HEAD`
 3. If **>10 commits since scan** OR **scan is older than 14 days**:
-   - Log: "Auto-refreshing scan docs (stale by {N} commits / {N} days)..."
-   - Re-run only the stale dimensions by spawning the relevant scan teammates:
-     - If `quality.md` stale → spawn quality teammate (model: sonnet)
-     - If `architecture.md` stale → spawn architecture teammate (model: haiku)
-     - If `security.md` stale → spawn security teammate (model: sonnet)
-   - Update `.gsd-t/scan/.cache.json` after refresh
+   - Log: "Auto-refreshing the tech-debt register (stale by {N} commits / {N} days)..."
+   - Re-run the scan by invoking the volume-scaled scan Workflow (`templates/workflows/gsd-t-scan.workflow.js`, same as `/gsd-t-scan`) — the probe re-slices the codebase and the finders refresh the register. This replaces the retired fixed-dimension teammate refresh.
 4. If fresh → proceed silently
 
 If `.gsd-t/scan/` doesn't exist at all → skip (no scan data to refresh).

package/commands/gsd-t-gap-analysis.md

@@ -14,14 +14,11 @@ If `.gsd-t/scan/.cache.json` exists:
 1. Read the cache and check `scannedAt` for each dimension
 2. Count commits since the scan: `git rev-list --count --after="{scannedAt}" HEAD`
 3. If **>10 commits since scan** OR **scan is older than 14 days**:
-   - Log: "Auto-refreshing scan docs (stale by {N} commits / {N} days)..."
-   - Re-run only the stale dimensions by spawning the relevant scan teammates:
-     - If `architecture.md` stale → spawn architecture teammate (model: haiku)
-     - If `quality.md` stale → spawn quality teammate (model: sonnet)
-   - Update `.gsd-t/scan/.cache.json` after refresh
+   - Log: "Auto-refreshing the tech-debt register (stale by {N} commits / {N} days)..."
+   - Re-run the scan by invoking the volume-scaled scan Workflow (`templates/workflows/gsd-t-scan.workflow.js`, same as `/gsd-t-scan`) — the probe re-slices the codebase and the finders refresh the register. This replaces the retired fixed-dimension teammate refresh.
 4. If fresh → proceed silently
 
-If `.gsd-t/scan/` doesn't exist at all → skip (no scan data to refresh).
+If no prior `.gsd-t/techdebt.md` exists at all → skip (no scan data to refresh).
 
 ## Step 1: Load Context
 

package/commands/gsd-t-help.md

@@ -196,9 +196,9 @@ Use these when user asks for help on a specific command:
 - **Use when**: Adding significant new functionality to existing project
 
 ### scan
-- **Summary**: Deep codebase analysis producing architecture docs and techdebt.md
+- **Summary**: Deep codebase analysis producing techdebt.md — volume-scaled native Workflow (`templates/workflows/gsd-t-scan.workflow.js`): a probe carves the codebase into per-area slices, one deep finder per slice (enumerate-don't-sample), a single verify pass drops false positives, synthesis dedups/re-ranks into the register. Finder count scales with codebase volume (1-3 slices tiny → 15-40 large), not a fixed 5.
 - **Auto-invoked**: No
-- **Creates**: `.gsd-t/scan/`, `.gsd-t/techdebt.md`
+- **Creates**: `.gsd-t/techdebt.md` (prior archived to `.gsd-t/techdebt_YYYY-MM-DD.md`), the `.gsd-t/scan/*.md` dimension files, the full living-doc set (`docs/architecture.md`, `docs/workflows.md`, `docs/infrastructure.md`, `docs/requirements.md`, `README.md` — merged not overwritten, via the M67 document phase), and the HTML scan report
 - **Use when**: Understanding an existing codebase or auditing technical debt
 
 ### gap-analysis

package/commands/gsd-t-init-scan-setup.md

@@ -101,16 +101,16 @@ If the project has source code, graph indexing (`.gsd-t/graph/`) will run as par
 
 ## Step 4: Deep Codebase Scan (gsd-t-scan)
 
-Execute the full scan workflow (same as `/gsd-t-scan`):
+Execute the full scan by invoking the volume-scaled scan Workflow (same as `/gsd-t-scan` — `templates/workflows/gsd-t-scan.workflow.js`):
 
-1. Scan across all dimensions: architecture, business rules, security, quality, contracts
-2. Build `.gsd-t/techdebt.md` register
-3. Cross-populate findings into living documents (docs/architecture.md, docs/workflows.md, docs/infrastructure.md, docs/requirements.md)
-4. Update README.md with discovered tech stack and setup info
+1. The volume-probe slices the codebase by area (scaling with volume, not a fixed dimension set); per-slice deep finders enumerate; a single verify pass drops false positives.
+2. Synthesis builds the `.gsd-t/techdebt.md` register and the `.gsd-t/scan/*.md` analysis files.
+3. Cross-populate findings into living documents (docs/architecture.md, docs/workflows.md, docs/infrastructure.md, docs/requirements.md) from the register + analysis files.
+4. Update README.md with discovered tech stack and setup info.
 
-Always use team mode for the scan unless the codebase is trivially small (< 5 files) or teams are explicitly disabled.
+The Workflow's fan-out scales with codebase volume automatically — no team-mode toggle.
 
-**If `.gsd-t/techdebt.md` already exists**: Append new findings, don't overwrite.
+**If `.gsd-t/techdebt.md` already exists**: the Workflow archives it to `.gsd-t/techdebt_YYYY-MM-DD.md` and continues TD numbering.
 
 ## Step 5: Generate Project CLAUDE.md (gsd-t-setup)