agentic-sdlc-wizard 1.43.0 → 1.44.1

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.
@@ -13,7 +13,7 @@
13
13
  "name": "sdlc-wizard",
14
14
  "source": ".",
15
15
  "description": "SDLC enforcement for AI agents — TDD, planning, self-review, CI shepherd",
16
- "version": "1.43.0",
16
+ "version": "1.44.1",
17
17
  "author": {
18
18
  "name": "Stefan Ayala"
19
19
  },
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "sdlc-wizard",
3
- "version": "1.43.0",
3
+ "version": "1.44.1",
4
4
  "description": "SDLC enforcement for AI agents — TDD, planning, self-review, CI shepherd",
5
5
  "author": {
6
6
  "name": "Stefan Ayala",
package/CHANGELOG.md CHANGED
@@ -4,6 +4,43 @@ All notable changes to the SDLC Wizard.
4
4
 
5
5
  > **Note:** This changelog is for humans to read. Don't manually apply these changes - just run the wizard ("Check for SDLC wizard updates") and it handles everything automatically.
6
6
 
7
+ ## [1.44.1] - 2026-04-27
8
+
9
+ ### Fixed
10
+
11
+ - **Autocompact compound-misconfig detection** — closes #207. Consumer reported autocompact firing at 12% context on a fresh `opus[1m]` session because they set BOTH `CLAUDE_AUTOCOMPACT_PCT_OVERRIDE=30` AND `CLAUDE_CODE_AUTO_COMPACT_WINDOW=400000` (a natural misreading of `CLAUDE_CODE_SDLC_WIZARD.md:1008`'s "or"-joined cell). The two compound: `30% × 400000 = 120000 tokens ≈ 12% of 1M`.
12
+ - **Doc fix**: `CLAUDE_CODE_SDLC_WIZARD.md` 1M-vs-200K table now writes `**OR** ... (pick one)` and adds a `> ⚠ Do NOT set both` callout that explains the compound math and points at the runtime detection.
13
+ - **Runtime detection**: `instructions-loaded-check.sh` (InstructionsLoaded hook) reads `.claude/settings.json` for both env vars, computes the effective trigger, and warns with the math when both are set — diagnosable from the warning alone.
14
+ - **Shipped skill drift**: `skills/sdlc/SKILL.md` was still calling `opus[1m]` the "default" (stale post-#198) AND repeating the same ambiguous "30 or 400000" wording it ships to consumers. Both fixed: opus[1m] now framed as opt-in with #198 reference; autocompact tuning line says "pick ONE of: ... OR ... (do NOT set both)".
15
+ - 4 new test-hooks tests (warns / silent on PCT-only / silent on WINDOW-only / shows effective trigger), 3 new test-doc-consistency tests (wizard doc + sdlc skill regression guards), size-cap test fixture extended to include the new branch (cap raised 1500 → 1700 to accommodate). Codex round 2 CERTIFIED 9/10 (round 1 surfaced the size-cap, shipped-skill drift, and InstructionsLoaded vs SessionStart wording — all fixed).
16
+
17
+ ### Files
18
+
19
+ - `hooks/instructions-loaded-check.sh` — new compound-misconfig detection block (single-line warning with full env var names + effective trigger math)
20
+ - `CLAUDE_CODE_SDLC_WIZARD.md` — line 1008 alternatives clarification + `> ⚠ Do NOT set both` callout
21
+ - `skills/sdlc/SKILL.md` — `opus[1m]` reframed `Default` → `Opt-in` (matches wizard doc post-#198); autocompact tuning line now warns against the compound config
22
+ - `tests/test-hooks.sh` — 4 new tests + size-cap fixture extended + cap raised
23
+ - `tests/test-doc-consistency.sh` — 3 new regression guards (wizard doc + sdlc skill)
24
+
25
+ ## [1.44.0] - 2026-04-27
26
+
27
+ ### Fixed
28
+
29
+ - **Install-path & cache hygiene** — closes #254 (P1+P2), #239, #238 filed by consumer codeguesser after upgrading 1.32.0 → 1.42.1.
30
+ - **#254 Bug 1 (P1)**: `cli/init.js` FILES list now ships `hooks/_find-sdlc-root.sh`. The helper is sourced by all 5 wizard hooks and provides `find_sdlc_root` + `dedupe_plugin_or_project`, but `npx ... init --force` previously didn't copy it. Result on every consumer: stderr noise (`_find-sdlc-root.sh: No such file or directory`, `dedupe_plugin_or_project: command not found`) and the SDLC walk-up logic (#171), silent-exit-for-non-SDLC-dirs (#173), and plugin/project dedupe were all silently dead.
31
+ - **#254 Bug 2 (P2)**: `init --force` now invalidates `~/.cache/sdlc-wizard/latest-version` (or `$SDLC_WIZARD_CACHE_DIR/latest-version` when overridden). Previously the cache held the pre-upgrade "latest" for up to 24h, producing reverse staleness nudges like `update available 1.42.1 → 1.41.1`. Console now prints a `BUST` line so the cache eviction is visible.
32
+ - **#254 Bug 2 / #239 (semver-direction)**: `instructions-loaded-check.sh` introduces `semver_lt` for numeric major/minor/patch comparison. Nudge gate switched from `LATEST_VERSION != INSTALLED_VERSION` (which fired on equality AND reverse direction) to `semver_lt INSTALLED LATEST` — so the nudge only fires when an actual upgrade is available. Cache reads gain a sanity check: cached "latest" must be >= installed, otherwise force a refetch.
33
+ - **#239 (npm failure surface)**: When `npm view` fails (e.g. EPERM from root-owned `~/.npm/_cacache/`) AND no usable cache exists, the hook now prints `npm view failed — version check unavailable (run 'npm view agentic-sdlc-wizard version' to debug)`. Previously the version-check block produced no output at all and the user had no signal that the staleness nudge was broken.
34
+ - **#238 (dual-channel ack sentinel)**: The "CLI skills + Claude plugin both present" warning gains an opt-in silence mechanism. Once the user runs `mkdir -p $DUAL_CACHE_DIR && touch $DUAL_ACK_FILE` (the exact command is printed inside the nudge), the warning stops firing. Sentinel lives at `$SDLC_WIZARD_CACHE_DIR/dual-channel-acknowledged` (defaults to `~/.cache/sdlc-wizard/dual-channel-acknowledged`). Removes the every-session noise that was training users to ignore all hook output.
35
+ - 8 new tests across `tests/test-cli.sh` (75 total, +3 for #254) and `tests/test-hooks.sh` (134 total, +5 for the four bugs + Codex round-1 cache-dir-absent regression test). Codex CERTIFIED 10/10 round 2.
36
+
37
+ ### Files
38
+
39
+ - `cli/init.js` — `FILES` list adds `_find-sdlc-root.sh`; new `invalidateVersionCache()` helper; `--force` path calls bust + logs `BUST` line
40
+ - `hooks/instructions-loaded-check.sh` — new `semver_lt()`; cache sanity check; npm-fail surface; semver-direction nudge gate; dual-channel ack sentinel
41
+ - `tests/test-cli.sh` — file-count test bumped 11→12; 3 new tests
42
+ - `tests/test-hooks.sh` — 5 new tests covering all 4 bugs + Codex cache-dir-absent regression
43
+
7
44
  ## [1.43.0] - 2026-04-27
8
45
 
9
46
  ### Added
@@ -1005,7 +1005,9 @@ Claude Code supports both 200K and 1M context windows. **`opus[1m]` is an opt-in
1005
1005
  | **Cost** | Standard pricing | Anthropic currently lists the 1M window at standard pricing across the full context for supported Opus/Sonnet models — **verify current rates at [docs.anthropic.com/pricing](https://docs.anthropic.com/)** before assuming no premium |
1006
1006
  | **Auto-mode** | **Enabled** — Claude Code chooses model per turn | **Disabled** — top-level `model` tells CC you've chosen explicitly |
1007
1007
  | **Auto-compact** | Default ~95% works well | Fires at ~76K by default ([issue #34332](https://github.com/anthropics/claude-code/issues/34332)) — pair with `CLAUDE_AUTOCOMPACT_PCT_OVERRIDE=30` |
1008
- | **Suggested override (if you pin)** | `CLAUDE_AUTOCOMPACT_PCT_OVERRIDE=75` | `CLAUDE_AUTOCOMPACT_PCT_OVERRIDE=30` or `CLAUDE_CODE_AUTO_COMPACT_WINDOW=400000` |
1008
+ | **Suggested override (if you pin)** | `CLAUDE_AUTOCOMPACT_PCT_OVERRIDE=75` | `CLAUDE_AUTOCOMPACT_PCT_OVERRIDE=30` **OR** `CLAUDE_CODE_AUTO_COMPACT_WINDOW=400000` (pick one) |
1009
+
1010
+ > **⚠ Do NOT set both.** `CLAUDE_AUTOCOMPACT_PCT_OVERRIDE` and `CLAUDE_CODE_AUTO_COMPACT_WINDOW` are alternatives, not complementary. Setting both compounds: `30% × 400000 = 120000` tokens, which is ~12% of a 1M window — autocompact fires almost immediately, destroying the headroom you opted in for. Pick one knob: either lower the trigger percentage (`PCT_OVERRIDE=30`) on the model's default 1M window, OR cap the working window (`AUTO_COMPACT_WINDOW=400000`) at the model's default 95% trigger. The `instructions-loaded-check.sh` `InstructionsLoaded` hook (fires on session start/resume) detects this misconfig and prints the effective trigger so you can debug from the warning alone (#207).
1009
1011
 
1010
1012
  **Why `opus[1m]` is opt-in (issue #198):**
1011
1013
  - **Pinning disables auto-mode.** Max-plan users pay for Claude Code's per-turn model selection (Sonnet for cheap tasks, Opus for hard ones, plus weekly-limit smoothing). A top-level `model` gives that up.
@@ -2918,7 +2920,7 @@ If deployment fails or post-deploy verification catches issues:
2918
2920
 
2919
2921
  **SDLC.md:**
2920
2922
  ```markdown
2921
- <!-- SDLC Wizard Version: 1.43.0 -->
2923
+ <!-- SDLC Wizard Version: 1.44.1 -->
2922
2924
  <!-- Setup Date: [DATE] -->
2923
2925
  <!-- Completed Steps: step-0.1, step-0.2, step-0.4, step-1, step-2, step-3, step-4, step-5, step-6, step-7, step-8, step-9 -->
2924
2926
  <!-- Git Workflow: [PRs or Solo] -->
@@ -3983,7 +3985,7 @@ Walk through updates? (y/n)
3983
3985
  Store wizard state in `SDLC.md` as metadata comments (invisible to readers, parseable by Claude):
3984
3986
 
3985
3987
  ```markdown
3986
- <!-- SDLC Wizard Version: 1.43.0 -->
3988
+ <!-- SDLC Wizard Version: 1.44.1 -->
3987
3989
  <!-- Setup Date: 2026-01-24 -->
3988
3990
  <!-- Completed Steps: step-0.1, step-0.2, step-1, step-2, step-3, step-4, step-5, step-6, step-7, step-8, step-9 -->
3989
3991
  <!-- Git Workflow: PRs -->
package/cli/init.js CHANGED
@@ -25,6 +25,10 @@ const FILES = [
25
25
  { src: 'hooks/instructions-loaded-check.sh', dest: '.claude/hooks/instructions-loaded-check.sh', executable: true, base: REPO_ROOT },
26
26
  { src: 'hooks/model-effort-check.sh', dest: '.claude/hooks/model-effort-check.sh', executable: true, base: REPO_ROOT },
27
27
  { src: 'hooks/precompact-seam-check.sh', dest: '.claude/hooks/precompact-seam-check.sh', executable: true, base: REPO_ROOT },
28
+ // #254 Bug 1: shared helper sourced by all hooks above. Must ship — without
29
+ // it, hooks emit "_find-sdlc-root.sh: No such file or directory" + the
30
+ // SDLC root walk-up logic is silently dead.
31
+ { src: 'hooks/_find-sdlc-root.sh', dest: '.claude/hooks/_find-sdlc-root.sh', base: REPO_ROOT },
28
32
  { src: 'skills/sdlc/SKILL.md', dest: '.claude/skills/sdlc/SKILL.md', base: REPO_ROOT },
29
33
  { src: 'skills/setup/SKILL.md', dest: '.claude/skills/setup/SKILL.md', base: REPO_ROOT },
30
34
  { src: 'skills/update/SKILL.md', dest: '.claude/skills/update/SKILL.md', base: REPO_ROOT },
@@ -235,6 +239,21 @@ function printOps(ops) {
235
239
  }
236
240
  }
237
241
 
242
+ // #254 Bug 2: clear the version cache on `init --force` so the
243
+ // instructions-loaded hook re-fetches latest from npm post-upgrade. Without
244
+ // this, the cache holds the pre-upgrade "latest" for up to 24h and the hook
245
+ // prints reverse nudges like "1.42.1 → 1.41.1".
246
+ function invalidateVersionCache({ dryRun }) {
247
+ const cacheDir = process.env.SDLC_WIZARD_CACHE_DIR
248
+ || path.join(os.homedir(), '.cache', 'sdlc-wizard');
249
+ const cacheFile = path.join(cacheDir, 'latest-version');
250
+ if (!fs.existsSync(cacheFile)) return false;
251
+ if (!dryRun) {
252
+ try { fs.rmSync(cacheFile, { force: true }); } catch (_) { /* best-effort */ }
253
+ }
254
+ return true;
255
+ }
256
+
238
257
  function init(targetDir, { force = false, dryRun = false } = {}) {
239
258
  if (!dryRun && !force) {
240
259
  const pluginPaths = detectPluginInstall();
@@ -291,6 +310,13 @@ function init(targetDir, { force = false, dryRun = false } = {}) {
291
310
  console.log(` ${GREEN}APPEND${RESET} .gitignore (${gitignoreAdds.join(', ')})`);
292
311
  }
293
312
 
313
+ // #254 Bug 2: bust the latest-version cache after an upgrade so the
314
+ // staleness nudge re-fetches a fresh value from npm. Only on --force,
315
+ // since that's the upgrade path.
316
+ if (force && invalidateVersionCache({ dryRun: false })) {
317
+ console.log(` ${CYAN}BUST${RESET} version cache (${path.join(process.env.SDLC_WIZARD_CACHE_DIR || path.join(os.homedir(), '.cache', 'sdlc-wizard'), 'latest-version')})`);
318
+ }
319
+
294
320
  console.log(`
295
321
  ${GREEN}SDLC Wizard installed successfully!${RESET}
296
322
 
@@ -44,14 +44,38 @@ fi
44
44
  SDLC_MD="$PROJECT_DIR/SDLC.md"
45
45
  # Strict x.y.z semver — rejects whitespace, "junk", "1.alpha.0", etc.
46
46
  SEMVER_RE='^[0-9]+\.[0-9]+\.[0-9]+$'
47
+
48
+ # semver_lt A B → exit 0 if A < B, exit 1 otherwise. Both args validated semver.
49
+ # Used so a stale-but-fresh cache from before an upgrade ("latest" < installed)
50
+ # doesn't fire a reverse-direction nudge (#254 Bug 2 / #239).
51
+ semver_lt() {
52
+ local a_major a_minor a_patch b_major b_minor b_patch
53
+ a_major=$(echo "$1" | awk -F. '{print $1+0}')
54
+ a_minor=$(echo "$1" | awk -F. '{print $2+0}')
55
+ a_patch=$(echo "$1" | awk -F. '{print $3+0}')
56
+ b_major=$(echo "$2" | awk -F. '{print $1+0}')
57
+ b_minor=$(echo "$2" | awk -F. '{print $2+0}')
58
+ b_patch=$(echo "$2" | awk -F. '{print $3+0}')
59
+ if [ "$a_major" -lt "$b_major" ]; then return 0; fi
60
+ if [ "$a_major" -gt "$b_major" ]; then return 1; fi
61
+ if [ "$a_minor" -lt "$b_minor" ]; then return 0; fi
62
+ if [ "$a_minor" -gt "$b_minor" ]; then return 1; fi
63
+ if [ "$a_patch" -lt "$b_patch" ]; then return 0; fi
64
+ return 1
65
+ }
66
+
47
67
  if [ -f "$SDLC_MD" ]; then
48
68
  INSTALLED_VERSION=$(grep -o 'SDLC Wizard Version: [0-9.]*' "$SDLC_MD" | head -1 | sed 's/SDLC Wizard Version: //')
49
69
  if [ -n "$INSTALLED_VERSION" ] && [[ "$INSTALLED_VERSION" =~ $SEMVER_RE ]]; then
50
70
  VERSION_CACHE_DIR="${SDLC_WIZARD_CACHE_DIR:-$HOME/.cache/sdlc-wizard}"
51
71
  VERSION_CACHE_FILE="$VERSION_CACHE_DIR/latest-version"
52
72
  LATEST_VERSION=""
73
+ NPM_FAILED=0
53
74
 
54
- # Use cache if present, <24h old, and contents are valid semver
75
+ # Use cache if present, <24h old, contents are valid semver, AND cached
76
+ # version is not strictly older than installed (#239: post-upgrade
77
+ # cache poison sanity check — if cache says "latest=1.41.1" but
78
+ # installed=1.43.0, the cache is poisoned, force a refetch).
55
79
  if [ -f "$VERSION_CACHE_FILE" ]; then
56
80
  if stat -f %m "$VERSION_CACHE_FILE" > /dev/null 2>&1; then
57
81
  CACHE_MTIME=$(stat -f %m "$VERSION_CACHE_FILE")
@@ -62,22 +86,37 @@ if [ -f "$SDLC_MD" ]; then
62
86
  if [ "$CACHE_AGE" -lt 86400 ]; then
63
87
  CACHE_CONTENT=$(cat "$VERSION_CACHE_FILE" 2>/dev/null) || CACHE_CONTENT=""
64
88
  if [[ "$CACHE_CONTENT" =~ $SEMVER_RE ]]; then
65
- LATEST_VERSION="$CACHE_CONTENT"
89
+ # Sanity check: cached "latest" must be >= installed.
90
+ if ! semver_lt "$CACHE_CONTENT" "$INSTALLED_VERSION"; then
91
+ LATEST_VERSION="$CACHE_CONTENT"
92
+ fi
66
93
  fi
67
94
  fi
68
95
  fi
69
96
 
70
- # Fetch from npm if cache miss / stale / malformed
97
+ # Fetch from npm if cache miss / stale / malformed / poisoned
71
98
  if [ -z "$LATEST_VERSION" ] && command -v npm > /dev/null 2>&1; then
72
- NPM_RESULT=$(npm view agentic-sdlc-wizard version 2>/dev/null) || NPM_RESULT=""
73
- if [[ "$NPM_RESULT" =~ $SEMVER_RE ]]; then
99
+ NPM_RESULT=$(npm view agentic-sdlc-wizard version 2>/dev/null)
100
+ NPM_RC=$?
101
+ if [ "$NPM_RC" -ne 0 ] || ! [[ "$NPM_RESULT" =~ $SEMVER_RE ]]; then
102
+ NPM_FAILED=1
103
+ else
74
104
  LATEST_VERSION="$NPM_RESULT"
75
105
  mkdir -p "$VERSION_CACHE_DIR" 2>/dev/null || true
76
106
  printf '%s' "$LATEST_VERSION" > "$VERSION_CACHE_FILE" 2>/dev/null || true
77
107
  fi
78
108
  fi
79
109
 
80
- if [ -n "$LATEST_VERSION" ] && [ "$LATEST_VERSION" != "$INSTALLED_VERSION" ]; then
110
+ # #239: surface npm failure once when cache miss + npm fails. Without
111
+ # this, the version-check block produces no output and the user has
112
+ # no way to know the staleness nudge is broken.
113
+ if [ -z "$LATEST_VERSION" ] && [ "$NPM_FAILED" -eq 1 ]; then
114
+ echo "npm view failed — version check unavailable (run 'npm view agentic-sdlc-wizard version' to debug)"
115
+ fi
116
+
117
+ # #254 Bug 2: only nudge when installed < latest (semver direction).
118
+ # Equality `!=` previously fired reverse nudges post-upgrade.
119
+ if [ -n "$LATEST_VERSION" ] && semver_lt "$INSTALLED_VERSION" "$LATEST_VERSION"; then
81
120
  # Minor-version delta: 1.25.0 vs 1.34.0 → 9
82
121
  INSTALLED_MINOR=$(echo "$INSTALLED_VERSION" | awk -F. '{print $2+0}')
83
122
  LATEST_MINOR=$(echo "$LATEST_VERSION" | awk -F. '{print $2+0}')
@@ -131,8 +170,35 @@ fi
131
170
  # this hook and model-effort-check.sh both fire on SessionStart, so two checks
132
171
  # would double-print the nudge and risk drifting out of sync.
133
172
 
134
- # Dual-channel install check (#181) nudge when CLI skills + Claude plugin both present
135
- if [ -d "$PROJECT_DIR/.claude/skills/update" ]; then
173
+ # Autocompact compound-misconfig check (#207). Setting BOTH
174
+ # CLAUDE_AUTOCOMPACT_PCT_OVERRIDE and CLAUDE_CODE_AUTO_COMPACT_WINDOW
175
+ # compounds — e.g. 30% × 400000 = 120000 token trigger, which on a 1M
176
+ # window fires at ~12% of context. The wizard doc lists them as
177
+ # alternatives ("PCT_OVERRIDE=30 OR AUTO_COMPACT_WINDOW=400000") but the
178
+ # "or" is easy to misread, and the consumer in #207 hit autocompact at
179
+ # 12% in a fresh session. Surface the misconfig with the effective
180
+ # trigger so it's diagnosable from the warning alone.
181
+ SETTINGS_JSON="$PROJECT_DIR/.claude/settings.json"
182
+ if [ -f "$SETTINGS_JSON" ]; then
183
+ AC_PCT=$(grep -o '"CLAUDE_AUTOCOMPACT_PCT_OVERRIDE"[[:space:]]*:[[:space:]]*"[0-9]*"' "$SETTINGS_JSON" \
184
+ | head -1 | sed 's/.*"\([0-9]*\)"$/\1/')
185
+ AC_WIN=$(grep -o '"CLAUDE_CODE_AUTO_COMPACT_WINDOW"[[:space:]]*:[[:space:]]*"[0-9]*"' "$SETTINGS_JSON" \
186
+ | head -1 | sed 's/.*"\([0-9]*\)"$/\1/')
187
+ if [ -n "$AC_PCT" ] && [ -n "$AC_WIN" ]; then
188
+ # Effective trigger = pct% of window (integer math; both pure digits per the regex).
189
+ AC_TRIGGER=$(( AC_PCT * AC_WIN / 100 ))
190
+ AC_PCT_OF_1M=$(( AC_TRIGGER * 100 / 1000000 ))
191
+ echo "WARNING: autocompact compound misconfig — CLAUDE_AUTOCOMPACT_PCT_OVERRIDE=${AC_PCT} AND CLAUDE_CODE_AUTO_COMPACT_WINDOW=${AC_WIN} both set in .claude/settings.json compound to ${AC_TRIGGER} tokens (~${AC_PCT_OF_1M}% of 1M). Pick one — see wizard doc '1M vs 200K' (#207)."
192
+ fi
193
+ fi
194
+
195
+ # Dual-channel install check (#181) — nudge when CLI skills + Claude plugin both present.
196
+ # #238: silenced once the user opts in via an ack sentinel. Sentinel is per-host
197
+ # (lives under $SDLC_WIZARD_CACHE_DIR/dual-channel-acknowledged) since the dual
198
+ # install is always a $HOME-scoped condition.
199
+ DUAL_CACHE_DIR="${SDLC_WIZARD_CACHE_DIR:-$HOME/.cache/sdlc-wizard}"
200
+ DUAL_ACK_FILE="$DUAL_CACHE_DIR/dual-channel-acknowledged"
201
+ if [ -d "$PROJECT_DIR/.claude/skills/update" ] && [ ! -f "$DUAL_ACK_FILE" ]; then
136
202
  for plugin_path in "$HOME/.claude/plugins-local/sdlc-wizard-wrap" "$HOME/.claude/plugins/cache/sdlc-wizard-local"; do
137
203
  if [ -d "$plugin_path" ]; then
138
204
  echo "WARNING: dual-install detected — CLI skills in .claude/skills/ AND Claude plugin at:"
@@ -140,6 +206,7 @@ if [ -d "$PROJECT_DIR/.claude/skills/update" ]; then
140
206
  echo " Duplicate /update-wizard commands come from running both channels. Pick one:"
141
207
  echo " - Keep plugin: remove .claude/skills/ from this project"
142
208
  echo " - Keep CLI: /plugin uninstall sdlc-wizard (or remove plugin dir)"
209
+ echo " - Keep both: mkdir -p \"$DUAL_CACHE_DIR\" && touch \"$DUAL_ACK_FILE\" (silences this nudge)"
143
210
  break
144
211
  fi
145
212
  done
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "agentic-sdlc-wizard",
3
- "version": "1.43.0",
3
+ "version": "1.44.1",
4
4
  "description": "SDLC enforcement for Claude Code — hooks, skills, and wizard setup in one command",
5
5
  "bin": {
6
6
  "sdlc-wizard": "cli/bin/sdlc-wizard.js"
@@ -170,16 +170,18 @@ When auto-approving, still announce your approach — just don't wait for approv
170
170
 
171
171
  ## Recommended Model
172
172
 
173
- **Default: `opus[1m]` (Opus 4.7 with 1M context window).** Run `/model opus[1m]` at the start of any non-trivial SDLC session.
173
+ **Opt-in: `opus[1m]` (Opus 4.7 with 1M context window).** Run `/model opus[1m]` at the start of any non-trivial SDLC session — but understand the tradeoff first (issue #198).
174
174
 
175
- **Why:**
175
+ **Why opt-in, not default:** A top-level `model` pin in `.claude/settings.json` disables Claude Code's per-turn model auto-selection. That's a real cost — Max-plan users pay for that auto-selection (Sonnet for cheap tasks, Opus for hard ones, plus weekly-limit smoothing). Pin only when you actually need the 1M headroom.
176
+
177
+ **Why pin to `opus[1m]` when you do opt in:**
176
178
  - SDLC sessions (plan → TDD → review → CI shepherd) accumulate context fast — plans, test output, diffs, review artifacts. 200K fills up before you're done.
177
179
  - Forced auto-compact mid-task loses your working state. Extra headroom is cheaper than re-reading files.
178
180
  - At time of writing, Anthropic lists 1M context at standard pricing for supported Opus/Sonnet models — verify current rates for your plan before relying on this.
179
181
 
180
182
  **Requires Claude Code v2.1.111+** for Opus 4.7.
181
183
 
182
- **Pair with `CLAUDE_AUTOCOMPACT_PCT_OVERRIDE=30`.** Without it, CC's default auto-compact on 1M fires at ~76K and defeats the purpose. The wizard's `cli/templates/settings.json` sets both defaults on install.
184
+ **Pair with `CLAUDE_AUTOCOMPACT_PCT_OVERRIDE=30`** when you opt in. Without it, CC's default auto-compact on 1M fires at ~76K and defeats the purpose. The setup wizard's Step 9.5 prompts to write both together (template ships with neither, opt-in only).
183
185
 
184
186
  **Fall back to `opus` (200K) only when:** your plan charges a premium for long-context prompts, the task is genuinely short (<30K), or team cost controls flag >200K prompts. See the "1M vs 200K Context Window" section in `CLAUDE_CODE_SDLC_WIZARD.md` for details.
185
187
 
@@ -606,7 +608,7 @@ CI passes -> Read review suggestions
606
608
  - `/clear` after 2+ failed corrections (context polluted — start fresh with better prompt)
607
609
  - Auto-compact fires at ~95% capacity — no manual management needed
608
610
  - After committing a PR, `/clear` before starting the next feature
609
- - **Autocompact tuning:** Set `CLAUDE_AUTOCOMPACT_PCT_OVERRIDE` to trigger compaction earlier (75% for 200K, 30% for 1M). On 1M models, the default fires at ~76K — set 30% or `CLAUDE_CODE_AUTO_COMPACT_WINDOW=400000` to use the full context window. See wizard doc "Autocompact Tuning" for full details
611
+ - **Autocompact tuning:** Set `CLAUDE_AUTOCOMPACT_PCT_OVERRIDE` to trigger compaction earlier (75% for 200K, 30% for 1M). On 1M models, the default fires at ~76K — pick ONE of: `CLAUDE_AUTOCOMPACT_PCT_OVERRIDE=30` **OR** `CLAUDE_CODE_AUTO_COMPACT_WINDOW=400000` (do NOT set both — they compound to 30% × 400K = 120K trigger ≈ 12% of 1M, which fires almost immediately, #207). See wizard doc "Autocompact Tuning" for full details
610
612
 
611
613
  **`--bare` mode (v2.1.81+):** `claude -p "prompt" --bare` skips ALL hooks, skills, LSP, and plugins. This is a complete wizard bypass — no SDLC enforcement, no TDD checks, no planning hooks. Use only for scripted headless calls (CI pipelines, automation) where you explicitly don't want wizard enforcement. Never use `--bare` for normal development work.
612
614
 
@@ -131,9 +131,11 @@ Parse all CHANGELOG entries between the user's installed version and the latest.
131
131
 
132
132
  ```
133
133
  Installed: 1.24.0
134
- Latest: 1.43.0
134
+ Latest: 1.44.1
135
135
 
136
136
  What changed:
137
+ - [1.44.1] Autocompact compound-misconfig detection — closes #207. Consumer reported autocompact firing at 12% context on a fresh opus[1m] session because they set BOTH `CLAUDE_AUTOCOMPACT_PCT_OVERRIDE=30` AND `CLAUDE_CODE_AUTO_COMPACT_WINDOW=400000` (a natural misreading of the "or"-joined override cell). The two compound: 30% × 400K = 120K trigger ≈ 12% of 1M. Three-pronged fix: (a) wizard doc clarifies alternatives with a `> ⚠ Do NOT set both` callout that shows the compound math; (b) `instructions-loaded-check.sh` (InstructionsLoaded hook) detects when both env vars are set in `.claude/settings.json`, computes the effective trigger, and warns with the math; (c) shipped `skills/sdlc/SKILL.md` was still calling opus[1m] the "default" (stale post-#198) AND repeating the same ambiguous wording — both fixed. 4 new hook tests + 3 new doc-consistency tests + size-cap fixture extended. Codex round 2 CERTIFIED 9/10.
138
+ - [1.44.0] Install-path & cache hygiene — closes #254, #239, #238 filed by consumer codeguesser after upgrading 1.32.0 → 1.42.1. (1) `cli/init.js` FILES list now ships `hooks/_find-sdlc-root.sh` — the helper sourced by all 5 hooks was missing from npm install path, so every session emitted `_find-sdlc-root.sh: No such file or directory` + `dedupe_plugin_or_project: command not found` and the SDLC walk-up logic was silently dead. (2) `init --force` now invalidates `~/.cache/sdlc-wizard/latest-version` so post-upgrade hooks re-fetch fresh values from npm instead of serving the pre-upgrade cache for 24h (which produced reverse "1.42.1 → 1.41.1" nudges). (3) instructions-loaded-check.sh now uses semver-direction comparison via new `semver_lt` function: nudge only fires when installed < latest, equality is silent, reverse direction is silent. Cache sanity-check rejects poisoned values (cached "latest" < installed → force refetch). (4) When `npm view` fails AND cache empty, hook now surfaces a one-line warning instead of going silent. (5) Dual-channel install nudge gains an opt-in silence sentinel — set via `mkdir -p $SDLC_WIZARD_CACHE_DIR && touch $SDLC_WIZARD_CACHE_DIR/dual-channel-acknowledged` (printed inside the nudge itself for discoverability). 8 new tests across test-cli.sh + test-hooks.sh, Codex CERTIFIED 10/10 round 2.
137
139
  - [1.43.0] Token-spike anomaly detection — ROADMAP #220 closure. New `hooks/token-spike-check.sh` (SessionStart, opt-in via `.metrics/`) ingests CC transcript usage (`input_tokens` / `output_tokens` / `cache_creation_input_tokens` / `cache_read_input_tokens`) into `.metrics/token-history.jsonl`, then warns when the last session's `costly_tokens` (input + cache_creation + output, excluding the cheap cache_read tier) exceeds median + 2σ over a rolling baseline. Catches silent CC-side caching regressions (per Anthropic's 2026-04-23 post-mortem) before they surface on the invoice. Uses MAD-based spread for the median metric so a single baseline outlier doesn't mask the next spike. 14 quality tests in `tests/test-token-spike.sh` (incl. malicious-transcript privacy probe, flat-baseline floor, median-vs-mean contrast, concurrent-ingest mkdir lock).
138
140
  - [1.42.2] PreCompact self-heal documented — ROADMAP #209 closure. Added `pr_number` opt-in to all 3 handoff template schemas (skill Step 1; wizard Round 1 + cross-model section). Self-heal logic shipped earlier with #229 but was undocumented, leaving the dead-code path. New `test_handoff_template_documents_pr_number` enforces template/doc parity. Together with #229 (mtime auto-expire) closes the "stuck PENDING handoff blocks /compact forever" footgun from both directions.
139
141
  - [1.42.1] CI hygiene fix — skip Claude PR review on wizard self-PRs. 7 self-PRs (v1.39.0–v1.42.0) had shipped with red `review` job (API canary firing on dead credit balance). Treated as "expected" but red normalizes red. Workflow `if:` now skips review on `BaseInfinity/claude-sdlc-wizard` repo only; consumer projects unaffected. 7 quality tests, mutation-verified (== inversion fails).