refactor: nuclear-review pass — hook-command unification, installer dedup, Codex-bridge hardening

CHANGELOG.md

@@ -4,6 +4,19 @@ All notable changes to cc-settings are documented here.
 
 > **Versioning** — cc-settings uses a single version number matching the installer (`src/setup.ts` `VERSION` constant, written to `~/.claude/.cc-settings-version` sentinel). Historical entries below 10.0 predate this unification; the jump from v8.x to v10.x in April 2026 realigned the product version with the installer version that was already ahead.
 
+## [Unreleased]
+
+### Changed
+
+- Codex bridge hardening: `--force` escape bypasses a sticky rate-limited/no-access verdict (which Codex emits even on auth mismatch); a fresh `available` verdict skips the per-call `codex login status` probe for 60s; timeouts now report partial output + a split/raise hint instead of an opaque exit code; `exec` appends `git status`/`diff --stat` so the changed files are always surfaced; `sanitizeOutput` strips ANSI and redacts secrets (`sk-`/`Bearer`/`Authorization`/`*_API_KEY|TOKEN|SECRET=`) on all returned output.
+- Hook tamper-defense: the three divergent "managed `~/.claude/src` hook command" classifiers (settings-merge, light-profile, audit-hooks) are unified into `src/lib/hook-command.ts`; the trusted regex is tightened to `(scripts|hooks)` only (drops `lib/`); hook-block traversal goes through a `HooksBlock`-schema-driven `iterCommandHooks`. `readFingerprint`/`readSrcManifest` now validate through zod, strip control chars from `installedAt`, and reject manifest keys with `..`/absolute paths.
+- Installer: `buildInstallPlan` is the single source of truth for install/prune footprint; `main()` split into `runMigrateOnly`/`runFullInstall`; the in-installer skill-prune loop removed. `managed-skills.ts` split into `ACTIVE_SKILLS` (completed — 7 missing current skills added: `codex`, `freeze`, `plan-ceo-review`, `proof-of-work`, `retro`, `review-batch`, `strategist`) + `TOMBSTONE_SKILLS`; `lint:skills` now asserts `ACTIVE_SKILLS` matches `skills/` on disk.
+- Settings merger is now pure: MCP-server preservation moved out of `settings-merge.ts` into `mcp.ts` `resolveMcpServers` (behavior unchanged).
+
+### Internal
+
+- `claude-audit.ts` split into `analyzeCommands` (model) + `renderAudit` (render); shared frontmatter-lint core extracted from `lint-skills`/`lint-knowledge`; `writeState` is now atomic (tmp+rename); `tool-cadence` `CounterState` rehydration goes through `normalizeCounterState`.
+
 ## [11.27.1] — 2026-06-19
 
 ### Fix — installer merger now deep-merges object config defaults

CLAUDE-FULL.md

@@ -137,4 +137,4 @@ Full threat model + remediation: see `SECURITY.md`.
 
 ### Skill library soft cap — 40
 
-Anthropic's Skills guide flags 20–50 skills as the point where the Skill selector starts struggling to read every description per turn. We sit at 34 cc-settings skills (Tier P1 cleanup May 2026: retired `audit`, `lenis`; merged `create-handoff`+`resume-handoff` → `handoff`, `discovery`+`prd` → `plan-feature`, `ask`+`premortem`+`compare-approaches` → `oracle`, `tdd` folded into `test`, `cc-sync`+`cc-update` → `cc`; folded `long-task` into `orchestrate`; demoted `write-a-skill` to `bun run new-skill` CLI; `nuclear-review` ported from Cursor team-kit May 2026; `share-learning` revived May 2026; `proof-of-work` + `review-batch` added May 2026 from the Orchestration Tax; `freeze` edit-scope lock ported from gstack June 2026). **Adding a new skill past 40 requires removing one** — re-evaluate `skills/` for consolidation candidates first. Validate the library with `bun run lint:skills`, which enforces the spec (kebab-case folders, frontmatter contract, no angle brackets, …) and surfaces the cap as a warning when crossed. Drift hides easily; let the linter catch it.
+Anthropic's Skills guide flags 20–50 skills as the point where the Skill selector starts struggling to read every description per turn. We sit at 35 cc-settings skills (Tier P1 cleanup May 2026: retired `audit`, `lenis`; merged `create-handoff`+`resume-handoff` → `handoff`, `discovery`+`prd` → `plan-feature`, `ask`+`premortem`+`compare-approaches` → `oracle`, `tdd` folded into `test`, `cc-sync`+`cc-update` → `cc`; folded `long-task` into `orchestrate`; demoted `write-a-skill` to `bun run new-skill` CLI; `nuclear-review` ported from Cursor team-kit May 2026; `share-learning` revived May 2026; `proof-of-work` + `review-batch` added May 2026 from the Orchestration Tax; `freeze` edit-scope lock ported from gstack June 2026). **Adding a new skill past 40 requires removing one** — re-evaluate `skills/` for consolidation candidates first. Validate the library with `bun run lint:skills`, which enforces the spec (kebab-case folders, frontmatter contract, no angle brackets, …) and surfaces the cap as a warning when crossed. Drift hides easily; let the linter catch it.

CLAUDE.md

@@ -17,7 +17,7 @@ for standards.
 
 ## About This Repo
 
-**TypeScript on Bun** (migrated from bash in April 2026; see git log for history). Runtime: `bun >=1.1.30`. Skills: 34 cc-settings skills (Tier P1 cleanup May 2026: retired 3, merged 7 pairs into 4, demoted 1 to CLI; `nuclear-review` ported from Cursor team-kit May 2026; share-learning revived May 2026; `freeze` edit-scope lock ported from gstack June 2026).
+**TypeScript on Bun** (migrated from bash in April 2026; see git log for history). Runtime: `bun >=1.1.30`. Skills: 35 cc-settings skills (Tier P1 cleanup May 2026: retired 3, merged 7 pairs into 4, demoted 1 to CLI; `nuclear-review` ported from Cursor team-kit May 2026; share-learning revived May 2026; `freeze` edit-scope lock ported from gstack June 2026).
 Deps: `zod`, `@inquirer/confirm`, `yaml`. Dev: `@biomejs/biome`, `typescript`,
 `@types/bun`.
 

MANUAL.md

@@ -98,7 +98,7 @@ You'll be asked to pick a starter — **satus** (Next.js, content sites) or **no
 | *"review my changes"* | `/review` — checks against TypeScript / React / a11y / performance rules |
 | *"audit my recent activity"* | `bun run claude-audit` — analyzes Bash command logs |
 
-**4. Don't memorize skills.** Just talk normally. The `Skill` tool auto-matches the right one. To see all 34 cc-settings skills, scroll to [All Skills](#all-skills) below. (Native Claude Code skills like `/loop`, `/schedule`, `/code-review`, `/review`, `/init`, `/security-review`, and any plugins like `sanity:*` or `vercel:*` load in addition — your session typically sees 60–80 skills total.)
+**4. Don't memorize skills.** Just talk normally. The `Skill` tool auto-matches the right one. To see all 35 cc-settings skills, scroll to [All Skills](#all-skills) below. (Native Claude Code skills like `/loop`, `/schedule`, `/code-review`, `/review`, `/init`, `/security-review`, and any plugins like `sanity:*` or `vercel:*` load in addition — your session typically sees 60–80 skills total.)
 
 **5. When something is unclear**, ask Claude directly: *"what skill handles X?"* or *"what just changed in cc-settings?"*. The setup is self-describing.
 

src/hooks/tool-cadence.ts

@@ -66,12 +66,14 @@ const FILE_EDIT_TOOLS: Record<string, string> = {
 interface CounterState {
   count: number;
   lastTool: string;
+  /** Debounce timestamp; genuinely absent until the first fire. */
   firedAt?: number;
-  files?: string[];
-  nudged?: boolean;
-  countAtNudge?: number;
-  filesAtNudge?: number;
-  escalated?: boolean;
+  // The rest are always present once normalizeCounterState() has run.
+  files: string[];
+  nudged: boolean;
+  countAtNudge: number;
+  filesAtNudge: number;
+  escalated: boolean;
 }
 
 type Payload = {
@@ -100,24 +102,41 @@ async function currentHead(cwd: string): Promise<string | undefined> {
   return out || undefined;
 }
 
+/** Validate and default every field of CounterState from an unknown raw value.
+ *  Replaces the previous inline re-hydration + `as number | undefined` cast. */
+function normalizeCounterState(raw: unknown): CounterState {
+  if (raw === null || typeof raw !== "object") {
+    return {
+      count: 0,
+      lastTool: "",
+      files: [],
+      nudged: false,
+      countAtNudge: 0,
+      filesAtNudge: 0,
+      escalated: false,
+    };
+  }
+  const r = raw as Record<string, unknown>;
+  return {
+    count: typeof r.count === "number" ? r.count : 0,
+    lastTool: typeof r.lastTool === "string" ? r.lastTool : "",
+    firedAt: typeof r.firedAt === "number" ? r.firedAt : undefined,
+    files: Array.isArray(r.files) ? (r.files as string[]) : [],
+    nudged: typeof r.nudged === "boolean" ? r.nudged : false,
+    countAtNudge: typeof r.countAtNudge === "number" ? r.countAtNudge : 0,
+    filesAtNudge: typeof r.filesAtNudge === "number" ? r.filesAtNudge : 0,
+    escalated: typeof r.escalated === "boolean" ? r.escalated : false,
+  };
+}
+
 // --- Branch 1: consecutive non-Agent call counter --------------------------
 
 async function parallelmaxBranch(
   toolName: string,
   toolInput: Payload["tool_input"],
 ): Promise<void> {
-  // Defensive defaults for old-shape state files.
-  const raw = await readState<CounterState>(COUNTER_STATE, { count: 0, lastTool: "" });
-  const state = {
-    count: raw.count ?? 0,
-    lastTool: raw.lastTool ?? "",
-    firedAt: raw.firedAt as number | undefined,
-    files: raw.files ?? ([] as string[]),
-    nudged: raw.nudged ?? false,
-    countAtNudge: raw.countAtNudge ?? 0,
-    filesAtNudge: raw.filesAtNudge ?? 0,
-    escalated: raw.escalated ?? false,
-  };
+  const raw = await readState<unknown>(COUNTER_STATE, null);
+  const state = normalizeCounterState(raw);
 
   if (toolName === "Agent") {
     // Reset the whole streak on delegation; preserve firedAt for debounce.