fix(ruvector-npm): accurate MCP tool count + reconcile tool lists (0.2.33) (#624)

* fix(ruvector-npm): accurate MCP tool count + reconcile tool lists (0.2.33)

The package advertised 103 MCP tools but the server registers 97, and the
README listed 12 tools (brain_agi_*, midstream_*) that don't exist. Three
independent sources had also drifted: server TOOLS array (97), the cli
'mcp tools' display list (91, missing the 6 decompile_* tools), and the
README count (103).

- README: 103 → 97; drop the 12 phantom tool bullets; point to
  'npx ruvector mcp tools' as the authoritative list
- cli.js 'mcp tools': add the missing 'decompile' group (6 tools) →
  now lists 97, matching the server
- mcp-server.js: report version from package.json instead of the
  hardcoded (and stale) '0.2.30'
- bump 0.2.32 → 0.2.33

All three sources now reconcile to 97. Full test suite passes
(73 + 7 + 2 + 8 + 2 across the publish-gate files).

Co-Authored-By: claude-flow <ruv@ruv.net>

* fix(ci): commit optimizer module, ignore RUSTSEC-2026-0190, ship src/optimizer

Root-causes three pre-existing CI failures on the ruvector npm package:

1. Optimizer test (Cannot find module src/optimizer/index.js): the module
   was never committed because .gitignore's bare 'index.js' rule (for napi
   artifacts) silently ignored this hand-written module. Add a negation
   exception (as already done for @ruvector/sona) and commit index.js —
   8 profiles, detectTaskType, applyProfile. Test: 135/135 pass.

2. cargo deny (advisories FAILED): RUSTSEC-2026-0190 (unsoundness in
   anyhow's Error::downcast_mut, anyhow 1.0.102, single version in tree).
   Added to deny.toml ignore list with justification + re-review date,
   consistent with the existing unsoundness exceptions — we never downcast
   anyhow::Error to a mismatched mutable type.

3. src/optimizer/ now ships (added to package.json files), so the published
   package's 'optimize' command works instead of graceful-degrading.

Pre-existing failures NOT addressed here (separate infra work; main is also
red): Functional smoke / Recall quality / Learning check fail because the
npm CI installs with --no-optional, so @ruvector/rvf + the native HNSW addon
are absent and the engine runs degraded. dependency-review is already
continue-on-error (awaiting a repo Dependency-Graph settings flip).

Co-Authored-By: claude-flow <ruv@ruv.net>

.gitignore

@@ -41,6 +41,10 @@ index.d.ts
 # (v3 emits no v2-style loader), which published an empty 0.1.6 (#516)
 !npm/packages/sona/index.js
 !npm/packages/sona/index.d.ts
+# ruvector optimizer is a hand-written module (not a napi artifact); the bare
+# `index.js` rule above wrongly ignored it, so it was never committed and CI's
+# optimizer test could not resolve it.
+!npm/packages/ruvector/src/optimizer/index.js
 
 # Environment variables and secrets
 .env

deny.toml

@@ -130,6 +130,14 @@ ignore = [
     # mmap loading and other crates; no fixed release on the 0.9 line yet.
     # We don't pass attacker-controlled offsets to memmap2. Re-review 2026-08-01.
     "RUSTSEC-2026-0186",
+
+    # anyhow 1.0.102 — unsoundness in `Error::downcast_mut()` (RUSTSEC-2026-0190,
+    # "unsound", not an exploitable vuln). Triggering it requires downcasting an
+    # `anyhow::Error` to a mismatched mutable type — a pattern we never use; our
+    # usage is construct/propagate/`Display`/read-only `downcast_ref`. Single
+    # version (1.0.102) in the tree. Bump to the patched anyhow once it ships and
+    # remove this entry. Re-review 2026-08-01.
+    "RUSTSEC-2026-0190",
 ]
 
 # ─────────────────────────────────────────────────────────────────────

npm/packages/ruvector/README.md

@@ -69,7 +69,7 @@ npx ruvector hooks init --pretrain --build-agents quality
 
 ### MCP Server Integration
 
-RuVector includes an MCP server for Claude Code with 103 tools:
+RuVector includes an MCP server for Claude Code with 97 tools:
 
 ```bash
 # Add to Claude Code
@@ -86,10 +86,11 @@ claude mcp add ruvector -- npx ruvector mcp start
 - `hooks_rag_context` — Semantic context retrieval
 - `hooks_attention_info`, `hooks_gnn_info` — Neural capabilities
 - `brain_search`, `brain_share`, `brain_status` — Shared brain knowledge
-- `brain_agi_status`, `brain_sona_stats`, `brain_temporal`, `brain_explore` — AGI diagnostics
-- `brain_midstream`, `brain_flags` — Midstream platform + feature flags
-- `midstream_status`, `midstream_attractor`, `midstream_scheduler` — Streaming analysis
-- `midstream_benchmark`, `midstream_search`, `midstream_health` — Latency benchmarks + health
+- `decompile_package`, `decompile_file`, `decompile_url` — Package/file decompilation + witness
+- `edge_status`, `edge_join`, `edge_balance` — Edge cluster coordination
+- `identity_generate`, `identity_show` — Agent identity
+
+See `npx ruvector mcp tools` for the full, authoritative tool list.
 
 **MCP tool-access policy (default-deny, ADR-256):** restrict the exposed/callable
 tool surface with environment variables — useful for least-privilege deployments.

npm/packages/ruvector/bin/cli.js

@@ -8200,7 +8200,7 @@ mcpCmd.command('info')
 
 mcpCmd.command('tools')
   .description('List all MCP tools with descriptions (JSON output)')
-  .option('--group <group>', 'Filter by group (hooks, workers, rvf, rvlite, brain, edge, identity)')
+  .option('--group <group>', 'Filter by group (hooks, workers, rvf, rvlite, brain, edge, identity, decompile)')
   .option('--json', 'JSON output')
   .action((opts) => {
     const tools = {
@@ -8323,6 +8323,14 @@ mcpCmd.command('tools')
         { name: 'identity_generate', desc: 'Generate new pi key' },
         { name: 'identity_show', desc: 'Show current identity' },
       ],
+      'decompile': [
+        { name: 'decompile_package', desc: 'Decompile an npm package + witness' },
+        { name: 'decompile_file', desc: 'Decompile a single file' },
+        { name: 'decompile_url', desc: 'Decompile from a URL' },
+        { name: 'decompile_diff', desc: 'Diff two decompiled artifacts' },
+        { name: 'decompile_search', desc: 'Search decompiled artifacts' },
+        { name: 'decompile_witness', desc: 'Verify decompile witness' },
+      ],
     };
 
     // Filter by group if specified

npm/packages/ruvector/bin/mcp-server.js

@@ -514,7 +514,7 @@ class Intelligence {
 const server = new Server(
   {
     name: 'ruvector',
-    version: '0.2.30',
+    version: require('../package.json').version,
   },
   {
     capabilities: {

npm/packages/ruvector/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ruvector",
-  "version": "0.2.32",
+  "version": "0.2.33",
   "description": "Self-learning vector database for Node.js \u2014 hybrid search, Graph RAG, FlashAttention-3, HNSW, 50+ attention mechanisms",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",
@@ -93,6 +93,7 @@
     "bin/",
     "dist/",
     "src/decompiler/",
+    "src/optimizer/",
     "wasm/",
     "README.md",
     "LICENSE"

npm/packages/ruvector/src/optimizer/index.js

@@ -0,0 +1,177 @@
+/**
+ * RVAgent Optimizer — Claude Code configuration profiles (ADR-139).
+ *
+ * Maps a task type (coding, research, quickfix, …) to an optimal set of
+ * `CLAUDE_CODE_*` environment variables and a permission mode. Profiles are
+ * derived from the decompiled Claude Code intelligence (see ADR-139) and are
+ * consumed by `npx ruvector optimize` and the settings generator.
+ *
+ * Public API:
+ *   - PERMISSION_MODES            — valid permission-mode strings
+ *   - listProfiles()              — array of profile names
+ *   - getProfile(name)            — { description, permissionMode, env } | null
+ *   - applyProfile(name)          — sets process.env, returns { applied, permissionMode } | null
+ *   - detectTaskType(prompt)      — infer a profile name from a free-text prompt
+ */
+
+'use strict';
+
+/** Permission modes understood by Claude Code's harness. */
+const PERMISSION_MODES = [
+  'default',
+  'acceptEdits',
+  'bypassPermissions',
+  'plan',
+  'dontAsk',
+  'auto',
+];
+
+/**
+ * Task profiles. Every env key MUST start with `CLAUDE_CODE_` and every
+ * `permissionMode` MUST be a member of {@link PERMISSION_MODES}.
+ */
+const PROFILES = {
+  coding: {
+    description: 'Implementation work — edits with checkpointing and thinking enabled',
+    permissionMode: 'acceptEdits',
+    env: {
+      CLAUDE_CODE_EFFORT_LEVEL: 'high',
+      CLAUDE_CODE_THINKING: '1',
+      CLAUDE_CODE_AUTO_COMPACT: '1',
+      CLAUDE_CODE_FILE_CHECKPOINTING: '1',
+    },
+  },
+  research: {
+    description: 'Investigation and analysis — read-heavy, no edits, deep reasoning',
+    permissionMode: 'default',
+    env: {
+      CLAUDE_CODE_EFFORT_LEVEL: 'high',
+      CLAUDE_CODE_THINKING: '1',
+      CLAUDE_CODE_AUTO_COMPACT: '1',
+    },
+  },
+  quickfix: {
+    description: 'Small, fast fixes — brief output, low effort, quick turnaround',
+    permissionMode: 'acceptEdits',
+    env: {
+      CLAUDE_CODE_BRIEF: '1',
+      CLAUDE_CODE_EFFORT_LEVEL: 'low',
+      CLAUDE_CODE_THINKING: '0',
+    },
+  },
+  planning: {
+    description: 'Architecture and design — plan mode, no edits until approved',
+    permissionMode: 'plan',
+    env: {
+      CLAUDE_CODE_EFFORT_LEVEL: 'high',
+      CLAUDE_CODE_THINKING: '1',
+    },
+  },
+  background: {
+    description: 'Long-running background/daemon work — autonomous, minimal prompts',
+    permissionMode: 'bypassPermissions',
+    env: {
+      CLAUDE_CODE_EFFORT_LEVEL: 'medium',
+      CLAUDE_CODE_AUTO_COMPACT: '1',
+      CLAUDE_CODE_PROMPT_SUGGESTION: '0',
+    },
+  },
+  swarm: {
+    description: 'Multi-agent swarm coordination — autonomous, hooks-driven',
+    permissionMode: 'bypassPermissions',
+    env: {
+      CLAUDE_CODE_EFFORT_LEVEL: 'medium',
+      CLAUDE_CODE_AUTO_COMPACT: '1',
+      CLAUDE_CODE_THINKING: '1',
+    },
+  },
+  review: {
+    description: 'Code review — read-only, no edits, thorough reasoning',
+    permissionMode: 'default',
+    env: {
+      CLAUDE_CODE_EFFORT_LEVEL: 'high',
+      CLAUDE_CODE_THINKING: '1',
+      CLAUDE_CODE_BRIEF: '0',
+    },
+  },
+  ci: {
+    description: 'CI / automation pipelines — fast, non-interactive, no suggestions',
+    permissionMode: 'dontAsk',
+    env: {
+      CLAUDE_CODE_FAST_MODE: '1',
+      CLAUDE_CODE_EFFORT_LEVEL: 'low',
+      CLAUDE_CODE_PROMPT_SUGGESTION: '0',
+    },
+  },
+};
+
+/**
+ * Keyword → profile rules, evaluated in order. The first rule whose regex
+ * matches the prompt wins, so more specific task types are listed first.
+ */
+const DETECTION_RULES = [
+  { type: 'quickfix', re: /\b(typo|quick\s?-?fix|one-?liner|small fix|rename)\b/i },
+  { type: 'ci', re: /\b(ci|cd|pipeline|github actions|workflow|deploy)\b/i },
+  { type: 'swarm', re: /\b(swarm|multi-?agent|coordinate|orchestrat\w*|hive)\b/i },
+  { type: 'background', re: /\b(background|daemon|monitor\w*|watch\b|long-?running)\b/i },
+  { type: 'review', re: /\b(review|pull request|\bpr\b|audit|critique)\b/i },
+  { type: 'planning', re: /\b(plan\w*|architect\w*|design|roadmap|strategy)\b/i },
+  { type: 'research', re: /\b(research|investigat\w*|explore|compare|survey)\b/i },
+  { type: 'coding', re: /\b(implement|code|build|add|write|refactor|create|fix|function|feature)\b/i },
+];
+
+/** List all available profile names. */
+function listProfiles() {
+  return Object.keys(PROFILES);
+}
+
+/** Return a profile by name, or `null` if it does not exist. */
+function getProfile(name) {
+  if (!name || !Object.prototype.hasOwnProperty.call(PROFILES, name)) {
+    return null;
+  }
+  return PROFILES[name];
+}
+
+/**
+ * Apply a profile's env vars to `process.env`. Returns `{ applied, permissionMode }`
+ * where `applied` is the map of env vars that were set, or `null` if the profile
+ * is unknown.
+ */
+function applyProfile(name) {
+  const profile = getProfile(name);
+  if (!profile) {
+    return null;
+  }
+  const applied = {};
+  for (const [key, val] of Object.entries(profile.env)) {
+    process.env[key] = val;
+    applied[key] = val;
+  }
+  return { applied, permissionMode: profile.permissionMode };
+}
+
+/**
+ * Infer the most likely task type from a free-text prompt. Defaults to
+ * `'coding'` for empty/unrecognised input.
+ */
+function detectTaskType(prompt) {
+  if (!prompt || typeof prompt !== 'string') {
+    return 'coding';
+  }
+  for (const rule of DETECTION_RULES) {
+    if (rule.re.test(prompt)) {
+      return rule.type;
+    }
+  }
+  return 'coding';
+}
+
+module.exports = {
+  PERMISSION_MODES,
+  PROFILES,
+  listProfiles,
+  getProfile,
+  applyProfile,
+  detectTaskType,
+};