its-magic 0.1.2-42 → 0.1.2-48

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -24,6 +24,28 @@ with pause/resume, decision gates, and persistent artifacts.
24
24
  - Built-in benchmarks (live, prompted, headless).
25
25
  - Multiplatform distribution (npm, Chocolatey, Homebrew).
26
26
 
27
+ <!-- readme-feature-coverage-catalog -->
28
+
29
+ ### Feature coverage catalog (US-0091)
30
+
31
+ - `/bin` — POSIX npm installer + Linux remote test targets (WSL / SSH / Docker) (`US-0084`).
32
+ - `/choco` — Configurable Multi-Target Release Publish with Confirmation Gate (`US-0054`).
33
+ - `/devops` — First-Class Bug Issue Workflow (Open/Closed) (`US-0079`).
34
+ - `/engineering` — Agent-Driven Codebase Map Bootstrap (`US-0082`).
35
+ - `/engineering` — Remote Runtime Connectivity Contract for QA/Release/Publish (`US-0064`).
36
+ - `/install` — Template/install payload omits intake gate scripts (`BUG-0001`).
37
+ - `/intake` — Optional Caveman-style input compression (safe file scope) (`US-0090`).
38
+ - `/lint` — CI/CD Workflows (`US-0007`).
39
+ - `/or` — Cursor Caveman mode (scratchpad-configurable terse responses) (`US-0089`).
40
+ - `/push` — Multiplatform Distribution (`US-0009`).
41
+ - `/run-tests` — Baseline Regression Cleanup for Installer and Version Sync Checks (`US-0074`).
42
+ - `/template` — End-to-End Lifecycle QA for `its-magic` Install/Upgrade/Clean (`US-0041`).
43
+ - `/upgrade` — Missing scripts still occur on install modes missing/upgrade (`BUG-0003`).
44
+ - `/usr` — Global Linux install fails: empty `install_include_paths` when manifest is CRLF (`BUG-0008`).
45
+ - `/workdir` — installer.sh fails in shell path with `set: Illegal option -` (`BUG-0004`).
46
+ - `MIGRATION` scratchpad flag — Smart Upgrade Mode (`US-0018`).
47
+ - `US-0016` scratchpad flag — Homebrew Version Sync (`US-0016`).
48
+
27
49
  ## Setup
28
50
 
29
51
  its-magic is an installer you run once per repo. It copies the AI dev team
@@ -947,6 +969,12 @@ Troubleshooting quick guide:
947
969
  - Fix JSON syntax and retry.
948
970
  - Secret-like inline value detected:
949
971
  - Replace literal secret with an env-var reference field.
972
+ - **CI still runs its-magic packaging jobs?** Your project received a pre-fix workflow.
973
+ Run **`its-magic --target <repo> --mode upgrade`** (or **`--mode clean`** then reinstall)
974
+ to refresh `.github/workflows/ci.yml` from the corrected template. After upgrade, GitHub
975
+ Actions should show only **`checks`** and **`auto-fix`** jobs — not `npm-test`,
976
+ `brew-test`, or `choco-test`. Fix applies to new installs/upgrades; stale repos heal on
977
+ next upgrade (**US-0018**).
950
978
 
951
979
  Team/local (recommended in `.cursor/scratchpad.local.md`):
952
980
 
@@ -1105,6 +1133,68 @@ push / PR ──> checks ──> PASS ──> done
1105
1133
  Auto-fix commits appear as `ci: auto-fix attempt N/3`. After 3 retries the
1106
1134
  workflow stops and points you to `scripts/validate-and-push` for local fixing.
1107
1135
 
1136
+ <!-- readme-feature-coverage-catalog -->
1137
+
1138
+ ### Feature coverage catalog (US-0091)
1139
+
1140
+ - `/acceptance` — Mandatory Intake Question Packs for First and Small Intakes (`US-0068`).
1141
+ - `/ask` — /ask Command: Context-Aware Questions Without Workflow (`US-0020`).
1142
+ - `/ask` — Context Compaction and Tiered Token-Cost Optimization Mode (`US-0053`).
1143
+ - `/auto` — Configurable Auto Phase Selection Policy (`US-0070`).
1144
+ - `/auto` — Continuous `/auto` Backlog-Drain Mode with Fine-Tune Switches (`US-0044`).
1145
+ - `/auto` — Fresh Subagent Context Per Phase and /auto Orchestration (`US-0023`).
1146
+ - `/auto` — Mid-Process `/auto` Continuation with Deterministic Resume Point (`US-0037`).
1147
+ - `/auto` — Strict Phase Role Enforcement in /auto Orchestration (`US-0069`).
1148
+ - `/auto` — Strict Runtime Proof for Per-Phase Subagent Isolation (`US-0056`).
1149
+ - `/auto` — Token-Cost Hardening for Orchestrated Runs (`US-0080`).
1150
+ - `/auto` — `/auto` continuous multi-phase loop + quiet drain (close one-phase-stop gap) (`US-0088`).
1151
+ - `/auto` — `/auto` executes phases without spawning required subagents (`BUG-0006`).
1152
+ - `/auto` — `/auto` explicit bug targeting (fix all OPEN bugs / fix `BUG-####`) (`US-0087`).
1153
+ - `/auto` — `/auto` fails with stale resume target after bug intake (`BUG-0005`).
1154
+ - `/check` — Optional Documentation Pack (Design Concept, CRS, Technical Spec) (`US-0031`).
1155
+ - `/confirmation` — Enforced Interactive Intake Question Evidence (`US-0078`).
1156
+ - `/connectivity` — Release Operator Run/Connect/Verify Hints Contract (`US-0067`).
1157
+ - `/decision` — Optional Fresh-Project ID Namespace Bootstrap (`US-0052`).
1158
+ - `/derived` — Deterministic Status Reconciliation Command (`US-0055`).
1159
+ - `/developer-dense` — Documentation Audience Profiles and Dual README Strategy (`US-0077`).
1160
+ - `/docs` — Installer-Owned `its_magic/` Folder for Framework Metadata (`US-0062`).
1161
+ - `/engineering` — Deterministic Context Slimming and Archive Enforcement Across Core Artifacts (`US-0072`).
1162
+ - `/engineering` — Deterministic State Hot-Surface Rollover and Archive Enforcement (`US-0060`).
1163
+ - `/engineering` — OS-Aware Runbook Command Auto-Bootstrap with Verified Quality Gates (`US-0063`).
1164
+ - `/flag` — Release Gate for Command/Flag Documentation Delta (`US-0030`).
1165
+ - `/intake` — Critical Evaluation in Intake and Architecture (`US-0021`).
1166
+ - `/intake` — Deterministic Intake Runtime Capability Guard and Single-Writer Drift Safety (`US-0059`).
1167
+ - `/intake` — Multi-Repo and Contract Compatibility Observability (`US-0034`).
1168
+ - `/intake` — intake evidence records asked questions that were never asked (`BUG-0007`).
1169
+ - `/integration` — Generated Test Scaffolding and Auto-Run Contract (`US-0066`).
1170
+ - `/managed` — Runtime QA Autopilot for Generated Projects (`US-0065`).
1171
+ - `/new` — First-Intake Full-Plan Coverage and Story-Map Gate (`US-0081`).
1172
+ - `/order` — Deterministic Artifact Ordering and Write Discipline (`US-0058`).
1173
+ - `/phases` — Cross-Phase Artifact Ownership Guard and Deterministic Archive Control (`US-0061`).
1174
+ - `/planning` — User-Visible Internal Metadata Sanitization Guard (`US-0071`).
1175
+ - `/product` — Backlog Reconciliation Gate for Released Sprints (`US-0043`).
1176
+ - `/push` — Phase-Triggered Sync Policy with Guarded Auto-Push (`US-0038`).
1177
+ - `/release` — Enforced Per-Phase Subagent Isolation with Audit Gate (`US-0048`).
1178
+ - `/release` — Legacy DONE-Story Acceptance/Traceability Backfill Guard (`US-0049`).
1179
+ - `/release` — Per-Sprint Release Notes and Release Queue Tracker (`US-0040`).
1180
+ - `/release` — Release Findings Artifact and Post-QA Issue Workflow (`US-0042`).
1181
+ - `/release` — Release Gate Tightening for Check-In Tests and QA/UAT Completion (`US-0039`).
1182
+ - `/remote` — Automation-driven remote execution selection (Docker / SSH / NL container intent) (`US-0086`).
1183
+ - `/remote` — Gitignored `.env` for remote and release connectivity (no AI read) (`US-0085`).
1184
+ - `/repetitive` — Delegable Intake Clarification Without Hard Blocks (`US-0083`).
1185
+ - `/research` — Knowledge Curation & Early Research (`US-0029`).
1186
+ - `/risk` — Intelligent Intake Decomposition and Risk-Aware PO Questioning (`US-0051`).
1187
+ - `/scratchpad` — Executable Scratchpad-Driven Sync and Auto-Push Wiring (`US-0076`).
1188
+ - `/scratchpad` — Scratchpad Delivery Simplification (Example-Only Install Policy) (`US-0073`).
1189
+ - `/scratchpad` — Upgrade Scratchpad Example–First Refresh (Fix Example Drift vs Materialized Baseline) (`US-0075`).
1190
+ - `/scratchpad` — Upgrade-Safe Scratchpad Example Refresh and Parity (`US-0057`).
1191
+ - `/sprint-plan` — Explicit `/sprint-plan --bulk` Mode (`US-0046`).
1192
+ - `/sprint-plan` — Sprint Sizing Rules and Configurable Sprint Planning (`US-0022`).
1193
+ - `/story` — Optional Feature User Guide Generation (`US-0032`).
1194
+ - `/uat` — UAT Artifact Lifecycle and Ownership (`US-0027`).
1195
+ - `SKILL` scratchpad flag — Skill and Templates (`US-0004`).
1196
+ - `US-0001` scratchpad flag — Core Workflow Commands (`US-0001`).
1197
+
1108
1198
  ## Walkthrough examples
1109
1199
 
1110
1200
  ### Example 1: New feature from idea
@@ -1240,6 +1330,35 @@ Reliable pattern:
1240
1330
  - `decisions/`: decision records.
1241
1331
  - `.github/workflows/`: CI/CD templates.
1242
1332
 
1333
+ <!-- readme-feature-coverage-catalog -->
1334
+
1335
+ ### Feature coverage catalog (US-0091)
1336
+
1337
+ - `/evidence` — Backlog-to-Sprint Traceability Contract (`US-0025`).
1338
+ - `/exit` — Milestone Lifecycle Definition and Exit Criteria (`US-0026`).
1339
+ - `/field` — Official Remote Config Template, Docs, and Fail-Fast Validation (`US-0036`).
1340
+ - `/installer` — Runbook Completion (`US-0015`).
1341
+ - `/intake` — Component-Scoped Execution Mode with Protection Guards (`US-0035`).
1342
+ - `/intake` — Configurable Guided Intake Behavior (`US-0033`).
1343
+ - `/map-codebase` — map-codebase does not write codebase-map in fresh repos (`BUG-0002`).
1344
+ - `/memory-audit` — Memory Drift Audit Command (`US-0024`).
1345
+ - `/product` — Canonical Story Status Source + Global Drift Guard (`US-0045`).
1346
+ - `/product` — Clean Install Hygiene and Complete Clean-Repo Coverage (`US-0050`).
1347
+ - `/security-review` — Security & Compliance Review Agent (`US-0028`).
1348
+ - `/skip` — Explicit Bulk Execute Orchestration Mode (`US-0047`).
1349
+ - `/strings` — Clean Placeholder Content from Templates and Active Files (`US-0019`).
1350
+ - `/write` — Artifact Templates and Starter Docs (`US-0006`).
1351
+ - `AUTO_FLOW_MODE` scratchpad flag — Automation Modes (`US-0011`).
1352
+ - `README` scratchpad flag — Voice Input Documentation (`US-0010`).
1353
+ - `TEAM_MODE` scratchpad flag — Team Mode (`US-0013`).
1354
+ - `US-0002` scratchpad flag — AI Behavior Rules (`US-0002`).
1355
+ - `US-0003` scratchpad flag — Subagent Definitions (`US-0003`).
1356
+ - `US-0005` scratchpad flag — Hook System (`US-0005`).
1357
+ - `US-0008` scratchpad flag — CLI Installer (`US-0008`).
1358
+ - `US-0012` scratchpad flag — Benchmark Suite (`US-0012`).
1359
+ - `US-0014` scratchpad flag — Quality Chain (3-Layer) (`US-0014`).
1360
+ - `US-0017` scratchpad flag — Template Drift Guard (`US-0017`).
1361
+
1243
1362
  ## Developer and release deep-dive
1244
1363
 
1245
1364
  ### CI/CD via runbook
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "its-magic",
3
- "version": "0.1.2-42",
3
+ "version": "0.1.2-48",
4
4
  "description": "its-magic - AI dev team workflow for Cursor.",
5
5
  "license": "MIT",
6
6
  "bin": {
@@ -1,5 +1,14 @@
1
1
  #!/usr/bin/env python3
2
- """Verify active vs template/scripts/ bytes match for DEC-0063 intake gate modules (BUG-0001)."""
2
+ """Verify active vs template/scripts/ bytes match for DEC-0063 intake gate modules (BUG-0001).
3
+
4
+ Scoped modes (DEC-0073 §10 / US-0090):
5
+ --scope=intake (default) DEC-0063 intake pair table.
6
+ --scope=caveman-compress DEC-0073 caveman input-compression pair table.
7
+ --scope=readme-feature-coverage DEC-0074 README feature-coverage pair table.
8
+ --scope=downstream-ci-guard DEC-0075 downstream CI guard script pair table.
9
+ --scope=us-0092 DEC-0078 full-autonomy outer driver + probe surfaces.
10
+ --scope=all union of all tables.
11
+ """
3
12
 
4
13
  from __future__ import annotations
5
14
 
@@ -7,7 +16,6 @@ import argparse
7
16
  import sys
8
17
  from pathlib import Path
9
18
 
10
- # Normative pairs: repo scripts/ (canonical dev) → template/scripts/ (packaged ship path).
11
19
  INTAKE_TEMPLATE_PAIRS: tuple[tuple[str, str], ...] = (
12
20
  ("scripts/intake_evidence_validate.py", "template/scripts/intake_evidence_validate.py"),
13
21
  ("scripts/intake_evidence_lib.py", "template/scripts/intake_evidence_lib.py"),
@@ -16,6 +24,86 @@ INTAKE_TEMPLATE_PAIRS: tuple[tuple[str, str], ...] = (
16
24
  ("scripts/check_intake_template_parity.py", "template/scripts/check_intake_template_parity.py"),
17
25
  )
18
26
 
27
+ # DEC-0073 §10 / US-0090 — Caveman input-compression surface pairs. Contents
28
+ # must be byte-identical between active and template paths; installer delivers
29
+ # template copies (BUG-0003 / DEC-0066).
30
+ CAVEMAN_COMPRESS_PAIRS: tuple[tuple[str, str], ...] = (
31
+ ("scripts/caveman_compress_input.py", "template/scripts/caveman_compress_input.py"),
32
+ ("docs/engineering/context/installer-owned-paths.manifest",
33
+ "template/docs/engineering/context/installer-owned-paths.manifest"),
34
+ ("docs/engineering/runbook.md", "template/docs/engineering/runbook.md"),
35
+ ("docs/engineering/auto-orchestration-reference.md",
36
+ "template/docs/engineering/auto-orchestration-reference.md"),
37
+ )
38
+
39
+ README_FEATURE_COVERAGE_PAIRS: tuple[tuple[str, str], ...] = (
40
+ (
41
+ "scripts/validate_readme_feature_coverage.py",
42
+ "template/scripts/validate_readme_feature_coverage.py",
43
+ ),
44
+ (
45
+ "scripts/readme_feature_coverage_lib.py",
46
+ "template/scripts/readme_feature_coverage_lib.py",
47
+ ),
48
+ (
49
+ "docs/engineering/context/readme-section-affinity.json",
50
+ "template/docs/engineering/context/readme-section-affinity.json",
51
+ ),
52
+ (".cursor/commands/release.md", "template/.cursor/commands/release.md"),
53
+ ("docs/engineering/runbook.md", "template/docs/engineering/runbook.md"),
54
+ (
55
+ "docs/engineering/context/installer-owned-paths.manifest",
56
+ "template/docs/engineering/context/installer-owned-paths.manifest",
57
+ ),
58
+ (
59
+ "scripts/check_intake_template_parity.py",
60
+ "template/scripts/check_intake_template_parity.py",
61
+ ),
62
+ )
63
+
64
+ US0092_PAIRS: tuple[tuple[str, str], ...] = (
65
+ ("scripts/auto_outer_driver.py", "template/scripts/auto_outer_driver.py"),
66
+ ("scripts/uat_probe_lib.py", "template/scripts/uat_probe_lib.py"),
67
+ (
68
+ "docs/engineering/context/installer-owned-paths.manifest",
69
+ "template/docs/engineering/context/installer-owned-paths.manifest",
70
+ ),
71
+ (".cursor/commands/auto.md", "template/.cursor/commands/auto.md"),
72
+ (".cursor/commands/verify-work.md", "template/.cursor/commands/verify-work.md"),
73
+ (".cursor/commands/qa.md", "template/.cursor/commands/qa.md"),
74
+ (
75
+ "docs/engineering/auto-orchestration-reference.md",
76
+ "template/docs/engineering/auto-orchestration-reference.md",
77
+ ),
78
+ ("docs/engineering/runbook.md", "template/docs/engineering/runbook.md"),
79
+ )
80
+
81
+ DOWNSTREAM_CI_GUARD_PAIRS: tuple[tuple[str, str], ...] = (
82
+ (
83
+ "scripts/check_downstream_ci_guard.py",
84
+ "template/scripts/check_downstream_ci_guard.py",
85
+ ),
86
+ (
87
+ "scripts/downstream_ci_guard_lib.py",
88
+ "template/scripts/downstream_ci_guard_lib.py",
89
+ ),
90
+ )
91
+
92
+ SCOPES: dict[str, tuple[tuple[str, str], ...]] = {
93
+ "intake": INTAKE_TEMPLATE_PAIRS,
94
+ "caveman-compress": CAVEMAN_COMPRESS_PAIRS,
95
+ "readme-feature-coverage": README_FEATURE_COVERAGE_PAIRS,
96
+ "downstream-ci-guard": DOWNSTREAM_CI_GUARD_PAIRS,
97
+ "us-0092": US0092_PAIRS,
98
+ "all": (
99
+ INTAKE_TEMPLATE_PAIRS
100
+ + CAVEMAN_COMPRESS_PAIRS
101
+ + README_FEATURE_COVERAGE_PAIRS
102
+ + DOWNSTREAM_CI_GUARD_PAIRS
103
+ + US0092_PAIRS
104
+ ),
105
+ }
106
+
19
107
 
20
108
  def main() -> int:
21
109
  p = argparse.ArgumentParser(description=__doc__)
@@ -25,10 +113,17 @@ def main() -> int:
25
113
  default=Path(__file__).resolve().parent.parent,
26
114
  help="Repository root",
27
115
  )
116
+ p.add_argument(
117
+ "--scope",
118
+ choices=sorted(SCOPES.keys()),
119
+ default="intake",
120
+ help="Parity pair table to verify.",
121
+ )
28
122
  args = p.parse_args()
29
123
  root: Path = args.repo
124
+ pairs = SCOPES[args.scope]
30
125
  failed = False
31
- for rel_active, rel_tpl in INTAKE_TEMPLATE_PAIRS:
126
+ for rel_active, rel_tpl in pairs:
32
127
  a = root / rel_active
33
128
  t = root / rel_tpl
34
129
  if not a.is_file() or not t.is_file():
@@ -45,7 +140,7 @@ def main() -> int:
45
140
  failed = True
46
141
  if failed:
47
142
  return 2
48
- print("[INTAKE_TEMPLATE_PARITY_OK]")
143
+ print(f"[INTAKE_TEMPLATE_PARITY_OK] scope={args.scope}")
49
144
  return 0
50
145
 
51
146
 
@@ -78,12 +78,22 @@ description: "its-magic architecture: define approach, risks, and decisions."
78
78
  - If `USER_GUIDE_MODE=0`, add no required user-guide steps or blocking checks (zero overhead).
79
79
  - If `USER_GUIDE_MODE=1`, reference canonical user-guide path and schema in
80
80
  architecture/state for in-scope feature stories; see runbook user-guide section.
81
- 9. Triad hot-surface gate (DEC-0054) when `docs/engineering/architecture.md` is
82
- mutated:
83
- - run `python scripts/enforce-triad-hot-surface.py --rollover` then `--check`
84
- from repository root,
85
- - on failure stop with `STATE_ARCHIVE_REQUIRED` or
86
- `ARTIFACT_HOT_SURFACE_OVERSIZE`,
81
+ 9. Triad hot-surface gate (DEC-0054; heading policy DEC-0076 / BUG-0010) when
82
+ `docs/engineering/architecture.md` is mutated:
83
+ - **Authoring mandate**: append new story sections as H1 `# US-xxxx` and new
84
+ bug sections as H1 `# BUG-xxxx` (not `## US-` / `## BUG-`).
85
+ - **Before** mutating the file, capture
86
+ `baseline_h2_count = count_h2_story_headings(architecture.md)` via
87
+ `python scripts/enforce-triad-hot-surface.py` (legacy `## US-xxxx` sections
88
+ remain rollover-visible; count must not increase during `/architecture`).
89
+ - Run `python scripts/enforce-triad-hot-surface.py --rollover` then `--check`
90
+ from repository root.
91
+ - Run heading policy check:
92
+ `python scripts/enforce-triad-hot-surface.py --check-arch-heading-policy
93
+ --baseline-h2-count <baseline_h2_count>`.
94
+ - On failure stop with `STATE_ARCHIVE_REQUIRED`, `ARTIFACT_HOT_SURFACE_OVERSIZE`,
95
+ or **`ARCH_STORY_HEADING_LEVEL_INVALID`** (non-suppressible `blocked` when H2
96
+ story-heading count increased),
87
97
  - preserve non-target history in archive packs only (never delete unrelated
88
98
  story sections without archival evidence).
89
99
  10. Codebase map lifecycle gate (US-0082 / DEC-0065) — before handoff to
@@ -71,6 +71,48 @@ Operators must follow the runbook recipe
71
71
  `stop_reason` vocabulary: `completed`, `decision_gate`, `missing_input`,
72
72
  `pause_request`, `loop_max`, `error`, `blocked`.
73
73
 
74
+ ## Full-autonomy mode + outer driver (US-0092 / DEC-0078)
75
+
76
+ **`AUTO_FLOW_MODE=full_autonomy`** (exact literal, default-off) enables the shipped
77
+ stdlib outer driver **`scripts/auto_outer_driver.py`**. The driver **loops hook
78
+ invocations** — spawn-only preserved (**BUG-0006**); it never performs phase-role work.
79
+
80
+ Operator recipe: set scratchpad keys → run
81
+ `python scripts/auto_outer_driver.py --repo .` once → interpret exit table in
82
+ **`docs/engineering/runbook.md`** § **Full-autonomy outer driver (US-0092)**.
83
+
84
+ **Drain-advance-without-pause**: with **`full_autonomy`** + **`AUTO_BACKLOG_DRAIN=1`**
85
+ (or bug-queue policy), segment completion schedules the next OPEN story/bug
86
+ **immediately** without operator re-`/auto`; **`resume_brief`** +
87
+ **`state.md`** refresh per **DEC-0069** at every boundary.
88
+
89
+ ### Full-autonomy stop matrix (US-0092)
90
+
91
+ **Invariant**: **`full_autonomy`** relaxes recoverable transient stops and operator
92
+ re-invocation, not governance gates. **`RELEASE_PUBLISH_MODE=auto`** remains
93
+ explicit opt-in for publish (unchanged default-off).
94
+
95
+ | Condition | US-0088 (all modes) | `full_autonomy` delta |
96
+ |-----------|---------------------|------------------------|
97
+ | Next phase, no hard stop | Continue inner `/auto` | Outer driver **re-invokes** when Cursor ends turn early |
98
+ | `decision_gate` | Hard stop | **No change — hard** |
99
+ | Unrecoverable `error` | Hard stop | **No change — hard** |
100
+ | Critical `missing_input` | Hard stop | **No change — hard** |
101
+ | Transient `missing_input` (recoverable) | Hard stop | **Relaxable** — bounded block-retry (`AUTO_BLOCK_RETRY_MAX`) |
102
+ | `pause_request` / `AUTO_PAUSE_REQUEST` | Hard stop | **No change — hard** |
103
+ | `loop_max` / `AUTO_LOOP_MAX_CYCLES` | Hard stop | **No change — hard** |
104
+ | `blocked` — transient/sync | Hard stop | **Relaxable** when ledger classifies recoverable |
105
+ | `blocked` — isolation / strict-proof / ownership | Hard stop | **No change — hard** |
106
+ | UAT/QA fail | Hard stop (operator) | **Relaxable** when `AUTO_IMPLEMENTATION_LOOP=1` |
107
+ | Segment complete + `AUTO_BACKLOG_DRAIN=1` | Advance (may need manual re-`/auto`) | **Drain-without-pause** — immediate next item |
108
+ | `BACKLOG_MAX_STORIES_REACHED` | Hard stop | **No change — hard** |
109
+ | `AUTO_SCHEDULER_CONFLICT` | Hard stop | **No change — hard** |
110
+ | Security deny (`.env`, intake evidence mutation) | Hard deny | **No change — hard** |
111
+
112
+ Block-retry ledger: append-only **`handoffs/auto_block_retry/<orchestrator_run_id>.jsonl`**
113
+ (names-only; cap exhaustion → exit **6** `BLOCK_RETRY_CAP_EXHAUSTED`). See architecture
114
+ **`# US-0092`** and **`docs/engineering/auto-orchestration-reference.md`**.
115
+
74
116
  ## Full specification (US-0080 / DEC-0062)
75
117
 
76
118
  Long prose, expanded mode semantics, and **Steps 1–13** detail live in
@@ -48,6 +48,14 @@ QA completion evidence is required before `/release` may proceed. Record finding
48
48
  `RELEASE_QA_BLOCKERS_OPEN`. When clear, handoff to `/verify-work` so release gate can
49
49
  verify no unresolved blockers.
50
50
 
51
+ ## Self-verify UAT probes (US-0092 / DEC-0078)
52
+
53
+ Share the **`scripts/uat_probe_lib.py`** resolver with **`/verify-work`**: derive
54
+ acceptance steps, execute probes where stack profile resolves, record evidence in
55
+ **`sprints/Sxxxx/uat.json`** `probe_results[]` and **`sprints/Sxxxx/qa-findings.md`**.
56
+ Fail closed with **`UAT_PROBE_UNRESOLVED`** (not PASS) when no probe maps.
57
+ Forbidden: auto-read **`.env`**, mutate intake evidence — **`UAT_PROBE_FORBIDDEN`**.
58
+
51
59
  ## Steps
52
60
  0. If `SECURITY_REVIEW=1`, verify `docs/engineering/security-review.md` exists
53
61
  and has no unresolved `critical` findings before proceeding. If unresolved
@@ -241,6 +241,17 @@ Guardrails:
241
241
  target-scoped repair and append entry to `docs/engineering/legacy-drift-audit.md`.
242
242
  - Behavior is deterministic and documented in runbook; do not mutate
243
243
  unrelated stories.
244
+ 3f. README feature coverage gate (US-0091 / DEC-0074):
245
+ - Read merged scratchpad `README_FEATURE_COVERAGE_ENFORCE` (default `0`).
246
+ - When `0`: skip with `skipped` evidence in `sprints/Sxxxx/release-findings.md`
247
+ § doc gates (grandfathering / migration pass).
248
+ - When `1`: run
249
+ `python scripts/validate_readme_feature_coverage.py --repo . --enforce`.
250
+ - On failure: emit `README_FEATURE_COVERAGE_BLOCKED` plus sub-codes on stderr;
251
+ remediation lists each missing id and target `root_h2` / `dev_h2` from
252
+ `docs/engineering/context/readme-section-affinity.json`.
253
+ - Active + `template/.cursor/commands/release.md` byte-identical step **3f**
254
+ block (full-file parity per US-0017).
244
255
  4. Verify UAT completeness (DEC-0009): confirm all sprint UAT artifacts (`uat.json`,
245
256
  `uat.md`) are in populated/verified state per DEC-0009. All steps must have
246
257
  recorded results. If any UAT is placeholder or incomplete, block release and
@@ -94,8 +94,23 @@ traceable for generated-project scope:
94
94
  `TEST_SCAFFOLD_GENERATION_FAILED` and require `/execute` or `/qa` rerun with
95
95
  deterministic evidence capture.
96
96
 
97
+ ## Self-verify UAT probes (US-0092 / DEC-0078)
98
+
99
+ When stack profile resolves, derive UAT steps from acceptance and execute probes via
100
+ the shared resolver **`scripts/uat_probe_lib.py`** (active + `template/scripts/` mirror).
101
+ Record evidence in **`sprints/Sxxxx/uat.json`** `probe_results[]` and
102
+ **`sprints/Sxxxx/qa-findings.md`** (path refs for stdout/stderr — no inline secrets).
103
+ **No silent PASS** — unresolvable steps must record fail-closed reason codes:
104
+
105
+ `UAT_PROBE_UNRESOLVED`, `UAT_STACK_PROFILE_UNKNOWN`, `UAT_PROBE_TIMEOUT`,
106
+ `UAT_PROBE_FAILED`, `UAT_PROBE_FORBIDDEN`, `UAT_PROBE_PASS`.
107
+
108
+ Probe catalog: **`build`**, **`test`**, **`api_health`**, **`process_health`**,
109
+ **`browser_smoke`**, **`cli_smoke`**, **`manual_operator`**. CLI:
110
+ `python scripts/uat_probe_lib.py --repo . --step "<acceptance step>" --report`.
111
+
97
112
  ## Steps
98
- 1. Convert acceptance criteria into testable UAT steps. Derive steps directly from the story's acceptance criteria in `docs/product/acceptance.md`. Each AC should map to at least one UAT step.
113
+ 1. Convert acceptance criteria into testable UAT steps. Derive steps directly from the story's acceptance criteria in `docs/product/acceptance.md`. Each AC should map to at least one UAT step. Run **`scripts/uat_probe_lib.py`** for each step where automation applies; record `probe_results[]` in **`uat.json`**.
99
114
  2. Populate UAT artifacts: write derived steps into `uat.json` (with description and result per step, accurate pass/fail counts) and `uat.md` (step list with results, summary section). Ensure UAT artifacts are in **populated** state per DEC-0009 — not placeholder.
100
115
  3. Record results and failures.
101
116
  4. Update state with pass/fail summary.
@@ -0,0 +1,184 @@
1
+ ---
2
+ description: "Caveman mode (US-0089) — optional response-side terse voice, default off"
3
+ globs: ["**/*"]
4
+ ---
5
+
6
+ # Caveman mode (US-0089)
7
+
8
+ This rule composes an optional **response-side** terse / imperative assistant
9
+ voice. It is **default off**. When `CAVEMAN_MODE` is unset or `0`, this rule
10
+ adds **zero** behavioral change and the assistant responds exactly as it did
11
+ pre-US-0089.
12
+
13
+ Inspired by JuliusBrussee/caveman (MIT). External reference only; not vendored.
14
+
15
+ ## Scratchpad gate (authoritative across subagent spawns)
16
+
17
+ Read the merged scratchpad (`.cursor/scratchpad.md`, then optional
18
+ `.cursor/scratchpad.local.md` overrides per **DEC-0055**) for these keys:
19
+
20
+ - `CAVEMAN_MODE=0|1` — default `0`. Absence is equivalent to `0`.
21
+ - `CAVEMAN_LEVEL=lite|full|ultra` — default empty.
22
+ - With `CAVEMAN_MODE=0`: inert regardless of value.
23
+ - With `CAVEMAN_MODE=1` and `CAVEMAN_LEVEL` empty: treat as `full`
24
+ (mid-tier default, matching upstream reference).
25
+ - With `CAVEMAN_MODE=1` and a listed level: apply that level.
26
+ - With `CAVEMAN_MODE=1` and any other non-empty value: fail closed with
27
+ deterministic reason code `CAVEMAN_LEVEL_UNKNOWN` and fall back to
28
+ pre-US-0089 voice (no terseness applied) while continuing the turn.
29
+ - `CAVEMAN_COMPRESS_INPUT=0|1` — **reserved for US-0090**; inert in US-0089;
30
+ no behavior until the compression story ships.
31
+ - `CAVEMAN_FILE_SCOPE=` — **reserved for US-0090**; inert in US-0089; no
32
+ behavior until the compression story ships.
33
+
34
+ Scratchpad values are **authoritative across subagent spawns**. Session
35
+ toggle phrases (see below) are overlays for the current conversation only
36
+ and do NOT persist across a fresh subagent context.
37
+
38
+ ## Literal-region invariant (9-zone hard MUST)
39
+
40
+ When `CAVEMAN_MODE=1`, the following output regions remain **byte-literal**
41
+ (no abbreviation, no rewording, no casing change). This rule is phrased as
42
+ **MUST**, not SHOULD:
43
+
44
+ 1. **Fenced code blocks** — both plain triple-backtick blocks and CODE
45
+ REFERENCE blocks of the form `startLine:endLine:filepath`. Content, line
46
+ numbers, filepath, language tags, and whitespace preserved.
47
+ 2. **File / path strings** — anything matching
48
+ `` `[\w./-]+\.(md|py|ps1|sh|json|mdc|ts|tsx|js|jsx|yml|yaml|toml|ini|env|example|manifest)` ``
49
+ or enclosed in backticks that points to a repo path. Preserve extension
50
+ and case.
51
+ 3. **AC checklist items** — bullets of form `- [ ]` or `- [x]` and their
52
+ entire text (AC-1..AC-N).
53
+ 4. **Reason codes** — `ALL_CAPS_WITH_UNDERSCORES` tokens such as
54
+ `PHASE_CONTEXT_ISOLATION_VIOLATION`, `RUNTIME_PROOF_MISSING`,
55
+ `AUTO_RESUME_ERROR`, `AUTO_SCHEDULER_CONFLICT`, `REMOTE_TARGET_UNKNOWN`,
56
+ `CAVEMAN_LEVEL_UNKNOWN`, `INTAKE_PERSISTENCE_BLOCKED`.
57
+ 5. **IDs** — `US-xxxx`, `DEC-xxxx`, `R-xxxx`, `BUG-####`, `S0xxx`, `T-xxx`.
58
+ 6. **Contract markers** — `[BUG_VALIDATION_OK]`,
59
+ `[INTAKE_EVIDENCE_VALIDATION_OK]`, `[SCRATCHPAD_PAIR_OK]`,
60
+ `[ARTIFACT_ORDERING_ANCHOR_AMBIGUOUS]`, `[CODEBASE_MAP_OK]`, and siblings.
61
+ 7. **Strict-proof tuple fields** — `orchestrator_run_id`, `runtime_proof_id`,
62
+ `proof_hash`, `proof_issued_at`, `proof_ttl_seconds`, `phase_id`, `role`
63
+ (DEC-0038).
64
+ 8. **Isolation evidence fields** — `fresh_context_marker`, `evidence_ref`,
65
+ `timestamp` (DEC-0029).
66
+ 9. **Commit / git refs** — quoted `git commit` messages, branch names, SHAs,
67
+ `HEAD`, tag names.
68
+
69
+ Forbidden "garbling" examples (never permitted even under `ultra`): dropping
70
+ the `US-` prefix from an ID, lowercasing a reason code, collapsing a fenced
71
+ code block into a bullet, truncating a path mid-filename, abbreviating a
72
+ proof hash.
73
+
74
+ ## Non-suppressible gate vocabulary (inherited from US-0088)
75
+
76
+ Caveman voice MUST NOT remove or abbreviate the following gate tokens — they
77
+ render byte-literal even under `CAVEMAN_LEVEL=ultra`:
78
+
79
+ - `decision_gate`
80
+ - `error`
81
+ - `pause`
82
+ - `loop_max`
83
+ - `blocked`
84
+ - `missing input`
85
+ - `[BUG_VALIDATION_OK]`
86
+ - `[INTAKE_EVIDENCE_VALIDATION_OK]`
87
+ - `[SCRATCHPAD_PAIR_OK]`
88
+
89
+ ## Operator toggle phrases (canonical set)
90
+
91
+ | Phrase | Effect |
92
+ |--------|--------|
93
+ | `caveman on` | Enable Caveman voice for the session (overlay). Effective from the next assistant turn. |
94
+ | `caveman off` | Disable Caveman voice for the session (overlay). Effective from the next assistant turn. |
95
+ | `stop caveman` | Alias for `caveman off`. |
96
+ | `normal mode` | Alias for `caveman off`. |
97
+ | `caveman: lite|full|ultra` | Set level for the session (implies `caveman on`). Effective from the next assistant turn. Accepts the three literal tokens `caveman: lite`, `caveman: full`, `caveman: ultra`. |
98
+
99
+ Determinism rules:
100
+
101
+ - Scratchpad `CAVEMAN_MODE` / `CAVEMAN_LEVEL` are **authoritative across
102
+ subagent spawns**; session toggles are overlays for the current
103
+ conversation only and do NOT persist across a fresh subagent context.
104
+ - Within a session, the **last explicit toggle wins**.
105
+ - A mid-turn toggle applies **from the next turn onward**; current-turn
106
+ machine-verifiable artifacts (gate messages, reason codes, strict-proof
107
+ tuples, isolation-evidence fields) remain literal regardless of the
108
+ toggle.
109
+ - Ambiguous phrases (`be caveman-lite`, `quiet caveman`, `cave man off`,
110
+ etc.) are **not recognized** — only the exact literal matches above fire.
111
+
112
+ ## Composition (non-substitution) with TOKEN_PROFILE
113
+
114
+ `TOKEN_PROFILE` controls context breadth. `CAVEMAN_MODE` controls reply
115
+ voice. Neither substitutes for the other; setting one does not change the
116
+ other. Combine freely.
117
+
118
+ ## Default-off invariant
119
+
120
+ With `CAVEMAN_MODE=0` or absent, this rule adds zero behavioral change:
121
+ existing command strings, gate ordering, spawn-only phrasing (BUG-0006),
122
+ contract-test token lists, `AUTO_QUIET` non-suppressible vocabulary
123
+ (US-0088), user-visible ID metadata (US-0071), strict-proof tuples
124
+ (DEC-0038), and isolation evidence fields (DEC-0029) all remain unchanged.
125
+
126
+ ## Non-goals (US-0089)
127
+
128
+ - No input-side file compression. `CAVEMAN_COMPRESS_INPUT` and
129
+ `CAVEMAN_FILE_SCOPE` are documented no-ops here; US-0090 owns that
130
+ vertical.
131
+ - No change to `TOKEN_PROFILE` (DEC-0035 / US-0080), context packs, archive
132
+ policy, or phase-context slimming.
133
+ - No rewrite of canonical artifacts (`docs/product/backlog.md`,
134
+ `docs/product/acceptance.md`, `docs/engineering/state.md` schema,
135
+ `docs/engineering/decisions.md`, DEC files).
136
+ - No new npm / Python dependencies. No `package.json` edit.
137
+ - No vendor plugin install. No package-manager install recipe is surfaced;
138
+ the JuliusBrussee/caveman reference is documentation-only.
139
+ - No change to spawn-only orchestration (US-0048 / DEC-0029 / BUG-0006),
140
+ strict runtime proof (DEC-0038), `AUTO_QUIET` non-suppressible list
141
+ (US-0088), or US-0071 visible-metadata rules.
142
+
143
+ ## Voice compression (when CAVEMAN_MODE=1)
144
+
145
+ ### Precedence
146
+
147
+ When `CAVEMAN_MODE=1`, voice-compression directives in **this rule** override
148
+ conflicting **user rule** prose-style guidance (e.g. "write complete sentences",
149
+ "blog-post quality") for **assistant reply voice only**; does **not** override
150
+ user rules for tool usage, git safety, security, or commit policy.
151
+
152
+ ### Intensity levels
153
+
154
+ | Level | Semantics | Example pattern |
155
+ |-------|-----------|-----------------|
156
+ | `lite` | Drop filler but keep grammatical sentences | Before: "I will now explain the orchestration gate because plan-verify requires it." After: "Orchestration gate required for plan-verify. Run gate check next." |
157
+ | `full` | Drop articles + fragments OK | Before: "The spawn-only orchestrator must dispatch a fresh subagent for each phase." After: "Spawn-only orchestrator dispatches fresh subagent per phase. Next: run `/execute`." Pattern: `[thing] [action] [reason]. [next step].` |
158
+ | `ultra` | Abbreviate ordinary prose words only; literals per §6 stub | Before: "The auth middleware returns HTTP 401 when the bearer token is absent from the Authorization header." After: "Auth middleware → 401. Bearer token absent. Check header." |
159
+
160
+ Goal: **token-saving terse/imperative prose** — **not** stereotypical caveman
161
+ roleplay.
162
+
163
+ ### Drop rules
164
+
165
+ Drop articles, filler, hedging, and pleasantries where level permits; keep
166
+ technical terms exact; fragments OK at `full` and `ultra`.
167
+
168
+ ### Auto-Clarity
169
+
170
+ Pause compression for **security** warnings, **destructive** operations, or
171
+ **ambiguous** compression — deliver the clear part in full prose, then resume
172
+ Caveman voice.
173
+
174
+ ### Persistence
175
+
176
+ Voice compression is active in **every response** while `CAVEMAN_MODE=1`;
177
+ turn off only via scratchpad keys or the documented toggle phrases above.
178
+
179
+ ### Ultra and literal regions
180
+
181
+ **Pointer stub** — defer to existing `## Literal-region invariant (9-zone hard
182
+ MUST)` above. Reason codes, IDs, paths, and proof tuples remain **byte-literal**.
183
+ At `ultra`, abbreviate ordinary prose words only — **never** abbreviate reason
184
+ codes or gate tokens. **No duplicate** 9-zone list here.