claude-flow 3.56.8 → 3.56.10

package/v3/@claude-flow/cli/dist/src/config/proven-config-refresh.js

@@ -84,6 +84,27 @@ export function currentInstallEnv(cwd = process.cwd()) {
     catch { /* none */ }
     return env;
 }
+/**
+ * #4176: is `cwd` a ruflo project? A bare `.claude/` is not enough: Claude Code
+ * itself creates `.claude/.cc-writes/` in whatever directory a Bash command ran
+ * in, so every subdirectory a session cd-ed into used to qualify and received a
+ * proven-config.json. A ruflo project carries `ruflo init`'s own markers: a
+ * `.claude/settings.json`, the `.claude/helpers/` directory, or an earlier
+ * adoption stamp.
+ */
+export function isRufloProject(cwd) {
+    try {
+        const claudeDir = path.join(cwd, '.claude');
+        if (!fs.statSync(claudeDir).isDirectory())
+            return false;
+        return fs.existsSync(path.join(claudeDir, 'settings.json'))
+            || fs.existsSync(path.join(claudeDir, 'helpers'))
+            || fs.existsSync(path.join(claudeDir, PROVEN_CONFIG_STAMP));
+    }
+    catch {
+        return false;
+    }
+}
 /**
  * The testable adoption core: given a signed champion and the local env, adopt
  * it into `cwd/.claude` iff it is newer than the stamp AND passes both gates.
@@ -92,8 +113,8 @@ export function currentInstallEnv(cwd = process.cwd()) {
 export function adoptSignedConfig(cwd, signed, env, opts = {}) {
     try {
         const claudeDir = path.join(cwd, '.claude');
-        if (!fs.existsSync(claudeDir))
-            return { adopted: false }; // not a ruflo project
+        if (!isRufloProject(cwd))
+            return { adopted: false }; // not a ruflo project (#4176)
         const championId = signed.manifest?.policy?.ref;
         if (!championId)
             return { adopted: false, skipped: 'manifest missing policy.ref' };
@@ -137,7 +158,7 @@ export function adoptSignedConfig(cwd, signed, env, opts = {}) {
  */
 export async function autoAdoptProvenConfigIfStale(cwd = process.cwd()) {
     try {
-        if (!fs.existsSync(path.join(cwd, '.claude')))
+        if (!isRufloProject(cwd))
             return { adopted: false };
         const src = findPackageProvenConfig();
         if (!src)

package/v3/@claude-flow/cli/dist/src/funnel/local-signals.js

@@ -34,7 +34,8 @@ export function getSecurityStatus(cwd = process.cwd()) {
         const rawFindings = scan.summary?.total ?? scan.totalFindings ?? scan.findings?.length ?? 0;
         const findings = Math.max(0, Number.isFinite(rawFindings) ? rawFindings : 0);
         return {
-            status: findings > 0 ? 'ISSUES' : 'CLEAN',
+            // #4124: a scan whose dependency phase could not run is not a clean scan.
+            status: findings > 0 ? 'ISSUES' : scan.incomplete === true ? 'PENDING' : 'CLEAN',
             findings,
             scannedAt: typeof scan.timestamp === 'string' ? scan.timestamp : undefined,
             cvesFixed: 0,

package/v3/@claude-flow/cli/dist/src/index.js

@@ -16,6 +16,7 @@ import { fileURLToPath } from 'url';
 import { dirname, join } from 'path';
 import { commandParser } from './parser.js';
 import { output } from './output.js';
+import { resolveTargetCommand } from './command-resolution.js';
 import { commands, getCommandsByCategory, getCommand, getCommandAsync, getCommandNames, getLazyCommandNames, hasCommand } from './commands/index.js';
 import { suggestCommand } from './suggest.js';
 import { runStartupUpdateCheck } from './update/index.js';
@@ -264,48 +265,11 @@ export class CLI {
                 this.output.writeln(this.output.dim(`  ${message}`));
                 process.exit(1);
             }
-            // Handle subcommand (supports nested subcommands)
-            let targetCommand = command;
-            let subcommandArgs = positional;
-            // Process command path (e.g., ['hooks', 'worker', 'list'])
-            // Note: When parser includes subcommand in commandPath, positional already excludes it
-            if (commandPath.length > 1 && command.subcommands) {
-                const subcommandName = commandPath[1];
-                const subcommand = command.subcommands.find(sc => sc.name === subcommandName || sc.aliases?.includes(subcommandName));
-                if (subcommand) {
-                    targetCommand = subcommand;
-                    // Parser already extracted subcommand from positional, so use as-is
-                    subcommandArgs = positional;
-                    // Check for nested subcommand (level 2)
-                    if (commandPath.length > 2 && subcommand.subcommands) {
-                        const nestedName = commandPath[2];
-                        const nestedSubcommand = subcommand.subcommands.find(sc => sc.name === nestedName || sc.aliases?.includes(nestedName));
-                        if (nestedSubcommand) {
-                            targetCommand = nestedSubcommand;
-                            // Parser already extracted nested subcommand too
-                            subcommandArgs = positional;
-                        }
-                    }
-                }
-            }
-            else if (positional.length > 0 && command.subcommands) {
-                // Check if first positional is a subcommand
-                const subcommandName = positional[0];
-                const subcommand = command.subcommands.find(sc => sc.name === subcommandName || sc.aliases?.includes(subcommandName));
-                if (subcommand) {
-                    targetCommand = subcommand;
-                    subcommandArgs = positional.slice(1);
-                    // Check for nested subcommand (level 2 from positional)
-                    if (subcommandArgs.length > 0 && subcommand.subcommands) {
-                        const nestedName = subcommandArgs[0];
-                        const nestedSubcommand = subcommand.subcommands.find(sc => sc.name === nestedName || sc.aliases?.includes(nestedName));
-                        if (nestedSubcommand) {
-                            targetCommand = nestedSubcommand;
-                            subcommandArgs = subcommandArgs.slice(1);
-                        }
-                    }
-                }
-            }
+            // Handle subcommands at any depth (#4157): walk the parsed command path, or the
+            // positional arguments when the path only names the top-level command.
+            const resolved = resolveTargetCommand(command, commandPath, positional);
+            const targetCommand = resolved.target;
+            const subcommandArgs = resolved.args;
             // Validate flags
             const validationErrors = this.parser.validateFlags(flags, targetCommand);
             if (validationErrors.length > 0) {

package/v3/@claude-flow/cli/dist/src/init/helpers-generator.js

@@ -478,7 +478,8 @@ module.exports = commands;
 }
 // Literal-word root guard mirrored in the mod and both shipped classic helpers.
 const ROOT_DELETE_CHECK_SOURCE = String.raw `function hasRootDelete(command, depth = 0) {
-  let word = '', quote = '', started = false, redirect = false
+  // mask has one letter per character of word: 'u' unquoted, 'q' quoted or escaped (only unquoted braces expand).
+  let word = '', mask = '', quote = '', started = false, redirect = false
   let ansiNul = false // a NUL inside $'...' ends that string, as in bash
   let inRm = false, optionsEnded = false, recursive = false, force = false, root = false
   const isRoot = (operand) => {
@@ -508,23 +509,92 @@ const ROOT_DELETE_CHECK_SOURCE = String.raw `function hasRootDelete(command, dep
     if (c === 'c' && at + 1 < s.length) return [String.fromCharCode(s.charCodeAt(at + 1) & 31), at + 1]
     return ['\\' + c, at]
   }
-  const finishWord = () => {
-    if (!started) return false
-    // Literal shell strings (e.g. sh -c 'rm -rf /') also carried the old guard.
-    // Bound rescanning to four levels; beyond that retain its conservative check.
-    if (/[\s;&|()\x60]/.test(word)) {
-      if (depth < 4 ? hasRootDelete(word, depth + 1) : word.includes('rm -rf /')) return true
+  // Past the rescan bound (or the brace budget), a string is judged by its tokens: an rm and a possible root operand
+  // anywhere in it deny. Quote characters are dropped first, so r"m" is still rm; a path token with braces may expand to /.
+  const looseRootDelete = (text) => {
+    const tokens = text.replace(/["'\\]/g, '').split(/[\s;&|()\x60$]+/)
+    return tokens.some(t => t === 'rm' || t.endsWith('/rm')) && tokens.some(t => isRoot(t) || (t.includes('/') && /[{}]/.test(t)))
+  }
+  // Bash brace expansion of one word ({a,b} lists, nested or glued to text), and nothing else: no {1..3} sequences, and
+  // nothing inside parameter braces after a dollar sign. Only unquoted braces and commas count. Iterative and bounded:
+  // past BRACE_WORDS words or BRACE_CHARS characters of work in this call it returns undefined.
+  const BRACE_WORDS = 4096
+  let braceChars = 1_000_000
+  // The leftmost expandable group: an unquoted {...} with an unquoted comma at its own level, outside any dollar-brace parameter. One pass.
+  const firstGroup = (text, marks) => {
+    const stack = []
+    let dollars = 0
+    let best
+    for (let k = 0; k < text.length; k++) {
+      if (marks[k] !== 'u') continue
+      const c = text[k]
+      if (c === '{') {
+        const dollar = k > 0 && text[k - 1] === '$' && marks[k - 1] === 'u'
+        if (dollar) dollars++
+        stack.push({ open: k, commas: [], dollar: dollar || dollars > 0 })
+      } else if (c === ',' && stack.length > 0) stack[stack.length - 1].commas.push(k)
+      else if (c === '}' && stack.length > 0) {
+        const top = stack.pop()
+        if (top.dollar && text[top.open - 1] === '$' && marks[top.open - 1] === 'u') dollars--
+        if (!top.dollar && top.commas.length > 0 && (best === undefined || top.open < best[0])) best = [top.open, top.commas, k]
+      }
+    }
+    return best
+  }
+  const braceExpand = (text, marks) => {
+    const out = []
+    const todo = [[text, marks]]
+    while (todo.length > 0) {
+      const [t, m] = todo.pop()
+      if ((braceChars -= t.length) < 0) return undefined
+      const group = firstGroup(t, m)
+      if (group === undefined) {
+        out.push(t)
+        if (out.length > BRACE_WORDS) return undefined
+        continue
+      }
+      const [open, commas, close] = group
+      const cuts = [open, ...commas, close]
+      // Pushed in reverse, so the words come out in bash's left-to-right order.
+      for (let c = cuts.length - 2; c >= 0; c--) {
+        const from = cuts[c] + 1, to = cuts[c + 1]
+        todo.push([t.slice(0, open) + t.slice(from, to) + t.slice(close + 1), m.slice(0, open) + m.slice(from, to) + m.slice(close + 1)])
+      }
+      if (todo.length > BRACE_WORDS) return undefined
     }
-    if (!inRm) inRm = word === 'rm' || word.endsWith('/rm')
-    else if (!optionsEnded && word === '--') optionsEnded = true
-    else if (!optionsEnded && word.startsWith('-')) {
+    return out
+  }
+  const takeWord = (w) => {
+    if (!inRm) inRm = w === 'rm' || w.endsWith('/rm')
+    else if (!optionsEnded && w === '--') optionsEnded = true
+    else if (!optionsEnded && w.startsWith('-')) {
       // GNU getopt accepts any unambiguous long-option prefix: --r, --recur, --forc.
-      recursive = recursive || (word.length > 2 && '--recursive'.startsWith(word)) || /^-[a-z]*r[a-z]*$/.test(word)
-      force = force || (word.length > 2 && '--force'.startsWith(word)) || /^-[a-z]*f[a-z]*$/.test(word)
-    } else root = root || isRoot(word)
-    word = ''; started = false
+      recursive = recursive || (w.length > 2 && '--recursive'.startsWith(w)) || (/^-[a-z]*$/.test(w) && w.includes('r'))
+      force = force || (w.length > 2 && '--force'.startsWith(w)) || (/^-[a-z]*$/.test(w) && w.includes('f'))
+    } else root = root || isRoot(w)
     return inRm && recursive && force && root
   }
+  const finishWord = () => {
+    if (!started) return false
+    const text = word, marks = mask
+    word = ''; mask = ''; started = false
+    // Literal shell strings (e.g. sh -c 'rm -rf /', or sh -c '{rm,-rf,/}') also carried the old guard.
+    // Bound rescanning to four levels; beyond that judge the string by its tokens.
+    if (/[\s;&|()\x60{]/.test(text)) {
+      if (depth < 4 ? hasRootDelete(text, depth + 1) : looseRootDelete(text)) return true
+    }
+    // An unquoted {a,b} becomes separate words, as bash expands it: rm -rf {/*,} is rm -rf /*.
+    const words = text.includes('{') ? braceExpand(text, marks) : [text]
+    if (words === undefined) {
+      // Too many alternatives to check: one of them may be the root (an rm in progress fails closed), or the
+      // list itself may spell the rm: judge its tokens.
+      if (inRm) root = true
+      if (looseRootDelete(text.replace(/[{},]/g, ' '))) return true
+      return takeWord(text)
+    }
+    for (const w of words) if (w !== '' && takeWord(w)) return true
+    return false
+  }
   // A command substitution's body, from just after its opening backtick. Bash removes a
   // backslash only before $, a backtick or a backslash (and " inside double quotes) and
   // drops backslash-newline; any other backslash stays for the nested scan.
@@ -538,7 +608,7 @@ const ROOT_DELETE_CHECK_SOURCE = String.raw `function hasRootDelete(command, dep
       }
       body += command[j]
     }
-    return [j, depth < 4 ? hasRootDelete(body, depth + 1) : body.includes('rm -rf /')]
+    return [j, depth < 4 ? hasRootDelete(body, depth + 1) : looseRootDelete(body)]
   }
   const finishCommand = () => {
     const denied = inRm && recursive && force && root
@@ -555,8 +625,8 @@ const ROOT_DELETE_CHECK_SOURCE = String.raw `function hasRootDelete(command, dep
         else if (char === '\\' && i + 1 < command.length) {
           const [text, last] = ansiEscape(command, i + 1)
           i = last
-          if (!ansiNul) { const nul = text.indexOf('\0'); if (nul < 0) word += text; else { word += text.slice(0, nul); ansiNul = true } }
-        } else if (!ansiNul) word += char
+          if (!ansiNul) { const nul = text.indexOf('\0'); const kept = nul < 0 ? text : text.slice(0, nul); word += kept; mask += 'q'.repeat(kept.length); if (nul >= 0) ansiNul = true }
+        } else if (!ansiNul) { word += char; mask += 'q' }
         continue
       }
       if (char === quote) quote = ''
@@ -566,20 +636,20 @@ const ROOT_DELETE_CHECK_SOURCE = String.raw `function hasRootDelete(command, dep
         // The substitution is gone from the word, as bash's empty output is (a quoted empty substitution glued to -rf leaves -rf).
         // A # right after it is still inside the word: a placeholder keeps a rescan (sh -c "...")
         // from reading it as the start of a comment.
-        if (command[end + 1] === '#') word += '\u0001'
+        if (command[end + 1] === '#') { word += '\u0001'; mask += 'q' }
         i = end
       }
       else if (quote === '"' && char === '\\' && i + 1 < command.length &&
         (command[i + 1] === '"' || command[i + 1] === '\\' || command[i + 1] === '$' ||
           command.charCodeAt(i + 1) === 96 || command[i + 1] === '\n')) {
         const next = command[++i]
-        if (next !== '\n') word += next
-      } else word += char
+        if (next !== '\n') { word += next; mask += 'q' }
+      } else { word += char; mask += 'q' }
       continue
     }
     if (char === '\\' && i + 1 < command.length) {
       const next = command[++i]
-      if (next !== '\n') { word += next; started = true }
+      if (next !== '\n') { word += next; mask += 'q'; started = true }
     } else if (char === '\x60' && command.indexOf('\x60', i + 1) > i) {
       // Command substitution: scan the body as its own command; the substitution
       // stays inside the enclosing word, so the rm being parsed keeps its state,
@@ -602,10 +672,26 @@ const ROOT_DELETE_CHECK_SOURCE = String.raw `function hasRootDelete(command, dep
       redirect = char === '<' || char === '>'
       if (!redirectionAmpersand && (char === '\n' || ';|&()'.includes(char)) && finishCommand()) return true
     } else {
-      word += char; started = true
+      word += char; mask += 'u'; started = true
     }
   }
   return finishWord() || finishCommand()
+}
+
+// A fork bomb's definition as bash reads it, whatever the spacing or the name: f(){ f|f& }, function f { f|f& },
+// function f(){ f|f& }. Defining one is the danger: the call may come anywhere later.
+function hasForkBomb(command) {
+  const s = command.replace(/\s+/g, '')
+  for (let brace = s.indexOf('{'); brace >= 0; brace = s.indexOf('{', brace + 1)) {
+    const end = s.startsWith('()', brace - 2) ? brace - 2 : brace
+    let start = end
+    while (start > 0 && end - start < 72 && /[a-z0-9_:.-]/.test(s[start - 1])) start--
+    const run = s.slice(start, end)
+    for (const name of run.startsWith('function') && run.length > 8 ? [run, run.slice(8)] : [run]) {
+      if (name && s.startsWith(name + '|' + name + '&}', brace + 1)) return true
+    }
+  }
+  return false
 }`;
 /**
  * Generate hook-handler.cjs (cross-platform hook dispatcher)
@@ -816,7 +902,7 @@ export function generateHookHandler() {
         "    var dangerous = ['rm -rf /', 'format c:', 'del /s /q c:\\\\', ':(){:|:&};:'];",
         ROOT_DELETE_CHECK_SOURCE.split('\n').map((line) => '    ' + line).join('\n'),
         '    for (var i = 0; i < dangerous.length; i++) {',
-        "      if (dangerous[i] === 'rm -rf /' ? hasRootDelete(cmd) : cmd.includes(dangerous[i])) {",
+        "      if (dangerous[i] === 'rm -rf /' ? hasRootDelete(cmd) : dangerous[i] === ':(){:|:&};:' ? hasForkBomb(cmd) : cmd.includes(dangerous[i])) {",
         "        console.error('[BLOCKED] Dangerous command detected: ' + dangerous[i]);",
         '        // Claude Code PreToolUse: exit 2 blocks execution; exit 1 is non-blocking.',
         '        process.exit(2);',

package/v3/@claude-flow/cli/dist/src/init/types.js

@@ -76,7 +76,7 @@ export const DEFAULT_INIT_OPTIONS = {
         github: true,
         flowNexus: false,
         browser: true,
-        v3: true,
+        v3: false, // ADR-128: opt-in like agents.v3; use --full or `init skills --v3`
         dualMode: false, // Optional: enable with --dual flag
         all: false,
     },

package/v3/@claude-flow/cli/dist/src/mcp-tools/agent-execute-core.d.ts

@@ -7,6 +7,7 @@
  * keeps the wire from agent_spawn → ProviderManager (real) in one
  * place rather than duplicated.
  */
+import { type AgentExecutionSample } from './agent-health-metrics.js';
 type ClaudeModel = 'haiku' | 'sonnet' | 'opus' | 'opus-4.7' | 'inherit';
 export interface AgentRecord {
     agentId: string;
@@ -32,6 +33,7 @@ export interface AgentRecord {
     /** ADR-148 phase 2 — concrete OpenRouter slug when provider='openrouter'. */
     openrouterModel?: string;
     lastResult?: Record<string, unknown>;
+    executionHistory?: AgentExecutionSample[];
 }
 export declare const DEFAULT_ANTHROPIC_MODEL = "claude-sonnet-5";
 export declare function modelRejectsSamplingParams(model: string): boolean;

package/v3/@claude-flow/cli/dist/src/mcp-tools/agent-execute-core.js

@@ -10,6 +10,7 @@
 import { existsSync, mkdirSync, readFileSync, writeFileSync } from 'node:fs';
 import { join } from 'node:path';
 import { getProjectCwd } from './types.js';
+import { appendAgentExecution, summarizeAgentExecutions } from './agent-health-metrics.js';
 import { configManager } from '../services/config-file-manager.js';
 const STORAGE_DIR = '.claude-flow';
 const AGENT_DIR = 'agents';
@@ -669,6 +670,20 @@ export async function executeAgentTask(input) {
             }
             if (out)
                 current.lastResult = out;
+            const previousMeasuredHealth = summarizeAgentExecutions(current.executionHistory).health;
+            const mayUpdateStoredHealth = previousMeasuredHealth === null
+                ? current.health === agent.health
+                : current.health === previousMeasuredHealth;
+            current.executionHistory = appendAgentExecution(current.executionHistory, {
+                at: new Date().toISOString(),
+                success: result.success,
+                durationMs: result.durationMs ?? Date.now() - startedAt,
+            });
+            // An explicit agent_update during the provider call still wins. The
+            // measured value remains available from executionHistory in agent_health.
+            if (mayUpdateStoredHealth) {
+                current.health = summarizeAgentExecutions(current.executionHistory).health ?? current.health;
+            }
             saveAgentStore(latest);
         };
         if (result.success) {

package/v3/@claude-flow/cli/dist/src/mcp-tools/agent-health-metrics.d.ts

@@ -0,0 +1,22 @@
+/** Bounded execution evidence for agent_health. Registry records are not processes. */
+export interface AgentExecutionSample {
+    at: string;
+    success: boolean;
+    durationMs: number;
+}
+export declare const AGENT_HEALTH_WINDOW = 20;
+export declare function appendAgentExecution(value: unknown, sample: AgentExecutionSample): AgentExecutionSample[];
+export declare function summarizeAgentExecutions(value: unknown): {
+    health: number | null;
+    errorRate: number | null;
+    sampleCount: number;
+    successes: number;
+    failures: number;
+    latency: {
+        avg: number | null;
+        p50: number | null;
+        p95: number | null;
+        p99: number | null;
+    };
+};
+//# sourceMappingURL=agent-health-metrics.d.ts.map

package/v3/@claude-flow/cli/dist/src/mcp-tools/agent-health-metrics.js

@@ -0,0 +1,33 @@
+export const AGENT_HEALTH_WINDOW = 20;
+function validSamples(value) {
+    if (!Array.isArray(value))
+        return [];
+    return value.filter((sample) => sample !== null && typeof sample === 'object' &&
+        typeof sample.at === 'string' && typeof sample.success === 'boolean' &&
+        typeof sample.durationMs === 'number' && Number.isFinite(sample.durationMs) && sample.durationMs >= 0).slice(-AGENT_HEALTH_WINDOW);
+}
+export function appendAgentExecution(value, sample) {
+    return [...validSamples(value), sample].slice(-AGENT_HEALTH_WINDOW);
+}
+export function summarizeAgentExecutions(value) {
+    const samples = validSamples(value);
+    const sampleCount = samples.length;
+    const successes = samples.filter(sample => sample.success).length;
+    const failures = sampleCount - successes;
+    const durations = samples.map(sample => sample.durationMs).sort((a, b) => a - b);
+    const percentile = (p) => sampleCount ? durations[Math.ceil(p * sampleCount) - 1] : null;
+    return {
+        health: sampleCount ? successes / sampleCount : null,
+        errorRate: sampleCount ? failures / sampleCount : null,
+        sampleCount,
+        successes,
+        failures,
+        latency: {
+            avg: sampleCount ? durations.reduce((sum, duration) => sum + duration, 0) / sampleCount : null,
+            p50: percentile(0.5),
+            p95: percentile(0.95),
+            p99: percentile(0.99),
+        },
+    };
+}
+//# sourceMappingURL=agent-health-metrics.js.map

package/v3/@claude-flow/cli/dist/src/mcp-tools/agent-tools.js

@@ -10,6 +10,7 @@ import { readRecordStore } from './record-store.js';
 import { getProjectCwd } from './types.js';
 import { validateIdentifier, validateText, validateAgentSpawn } from './validate-input.js';
 import { executeAgentTask } from './agent-execute-core.js';
+import { AGENT_HEALTH_WINDOW, summarizeAgentExecutions } from './agent-health-metrics.js';
 import { pheromoneAgentEligibility } from './swarm-tools.js';
 // Storage paths
 const STORAGE_DIR = '.claude-flow';
@@ -691,7 +692,7 @@ export const agentTools = [
     },
     {
         name: 'agent_health',
-        description: 'Compute an agent\'s rolling health score (0-1) from recent task success ratio + response-latency p50/p95 + error rate. Use when native Task tool is wrong because you\'re running a long-lived agent (autonomous loop / hive-mind worker / federation peer) and need to detect degradation before the breaker trips it. For one-shot Task invocations there is no history to score. Pair with hooks_post-task so the scores stay current.',
+        description: 'Report measured agent_execute provider-call outcomes over the last 20 calls: API success-ratio health (0-1), error rate, and response-latency p50/p95. This does not measure task correctness. Agents with no observed calls have unknown health; registration is not proof of a running process. Native Task and hooks_post-task outcomes are not recorded here. Use when you need measured reliability for agents driven through agent_execute; native Task results are invisible to this tool, so inspect those directly.',
         category: 'agent',
         inputSchema: {
             type: 'object',
@@ -708,58 +709,80 @@ export const agentTools = [
             }
             const store = loadAgentStore();
             const agents = Object.values(store.agents).filter(a => a.status !== 'terminated');
-            const threshold = input.threshold || 0.5;
+            const threshold = input.threshold === undefined ? 0.5 : input.threshold;
+            if (!Number.isFinite(threshold) || threshold < 0 || threshold > 1) {
+                return { error: 'threshold must be a finite number between 0 and 1' };
+            }
+            const healthStatus = (health) => health === null ? 'unknown' : health >= threshold ? 'healthy' : health >= 0.3 ? 'degraded' : 'unhealthy';
             if (input.agentId) {
                 const agent = store.agents[input.agentId];
                 if (agent) {
+                    const metrics = summarizeAgentExecutions(agent.executionHistory);
                     return {
                         agentId: agent.agentId,
-                        health: agent.health,
+                        ...metrics,
+                        healthStatus: healthStatus(metrics.health),
                         status: agent.status,
-                        healthy: agent.health >= threshold,
+                        healthy: metrics.health === null ? null : metrics.health >= threshold,
                         taskCount: agent.taskCount,
-                        uptime: Date.now() - new Date(agent.createdAt).getTime(),
+                        uptime: null,
+                        registryAgeMs: Date.now() - new Date(agent.createdAt).getTime(),
+                        windowSize: AGENT_HEALTH_WINDOW,
+                        source: 'agent_execute',
                     };
                 }
                 return { agentId: input.agentId, error: 'Agent not found' };
             }
-            const healthyAgents = agents.filter(a => a.health >= threshold);
-            const degradedAgents = agents.filter(a => a.health >= 0.3 && a.health < threshold);
-            const unhealthyAgents = agents.filter(a => a.health < 0.3);
-            const avgHealth = agents.length > 0 ? agents.reduce((sum, a) => sum + a.health, 0) / agents.length : 1;
+            const observed = agents.map(agent => ({ agent, metrics: summarizeAgentExecutions(agent.executionHistory) }));
+            const healthyAgents = observed.filter(({ metrics }) => healthStatus(metrics.health) === 'healthy');
+            const degradedAgents = observed.filter(({ metrics }) => healthStatus(metrics.health) === 'degraded');
+            const unhealthyAgents = observed.filter(({ metrics }) => healthStatus(metrics.health) === 'unhealthy');
+            const measuredAgents = observed.filter(({ metrics }) => metrics.health !== null);
+            const unknownCount = agents.length - measuredAgents.length;
+            const avgHealth = measuredAgents.length > 0
+                ? measuredAgents.reduce((sum, { metrics }) => sum + metrics.health, 0) / measuredAgents.length
+                : null;
             return {
                 // CLI expected fields
-                agents: agents.map(a => {
-                    const uptime = Date.now() - new Date(a.createdAt).getTime();
+                agents: observed.map(({ agent, metrics }) => {
                     return {
-                        id: a.agentId,
-                        type: a.agentType,
-                        health: a.health >= threshold ? 'healthy' : (a.health >= 0.3 ? 'degraded' : 'unhealthy'),
-                        uptime,
-                        tasks: { active: a.taskCount > 0 ? 1 : 0, queued: 0, completed: a.taskCount, failed: 0 },
-                        _note: 'Per-agent OS metrics not available — use system_metrics for real CPU/memory',
+                        id: agent.agentId,
+                        type: agent.agentType,
+                        health: healthStatus(metrics.health),
+                        score: metrics.health,
+                        sampleCount: metrics.sampleCount,
+                        errorRate: metrics.errorRate,
+                        latency: metrics.latency,
+                        errors: { count: metrics.failures },
+                        uptime: null,
+                        registryAgeMs: Date.now() - new Date(agent.createdAt).getTime(),
+                        tasks: { active: agent.status === 'busy' ? 1 : 0, queued: null, completed: metrics.successes, failed: metrics.failures },
+                        _note: `Provider-call outcomes and latency cover only the last ${AGENT_HEALTH_WINDOW} agent_execute calls; task correctness, process liveness and CPU/memory are not observed`,
                     };
                 }),
                 overall: {
                     healthy: healthyAgents.length,
                     degraded: degradedAgents.length,
                     unhealthy: unhealthyAgents.length,
+                    unknown: unknownCount,
                     cpu: null,
                     memory: null,
                     _note: 'Per-agent CPU/memory not available — use system_metrics for real OS-level stats',
-                    score: Math.round(avgHealth * 100),
+                    score: avgHealth === null ? null : Math.round(avgHealth * 100),
                     issues: unhealthyAgents.length,
                 },
                 // Additional fields
                 total: agents.length,
+                observedCount: measuredAgents.length,
+                unknownCount,
                 healthyCount: healthyAgents.length,
                 unhealthyCount: unhealthyAgents.length,
                 threshold,
                 avgHealth,
-                unhealthyAgents: unhealthyAgents.map(a => ({
-                    agentId: a.agentId,
-                    health: a.health,
-                    status: a.status,
+                unhealthyAgents: unhealthyAgents.map(({ agent, metrics }) => ({
+                    agentId: agent.agentId,
+                    health: metrics.health,
+                    status: agent.status,
                 })),
             };
         },

package/v3/@claude-flow/cli/dist/src/mcp-tools/agentbbs-federation.d.ts

@@ -50,7 +50,7 @@
  *
  * @module @claude-flow/cli/mcp-tools/agentbbs-federation
  */
-import { type Server } from 'node:http';
+import type { Server } from 'node:http';
 /** Max envelopes accepted from one peer in one sync. */
 export declare const MAX_ENVELOPES_PER_SYNC = 5000;
 /** Max serialized bytes for a single envelope. */

package/v3/@claude-flow/cli/dist/src/mcp-tools/agentbbs-federation.js

@@ -53,7 +53,12 @@
 import { existsSync, mkdirSync, readFileSync, writeFileSync, appendFileSync, chmodSync } from 'node:fs';
 import { join } from 'node:path';
 import { createHash, randomBytes, timingSafeEqual } from 'node:crypto';
-import { createServer } from 'node:http';
+import { createRequire } from 'node:module';
+// Lazy, CommonJS-style load of node:http: an ESM `import … from 'node:http'`
+// materialises every export of the builtin, which trips http's lazy undici
+// getters (~2.5 MB heap / ~6 MB RSS) at MCP startup even when no server is
+// ever started. require() of the builtin does not touch those getters.
+const loadHttp = () => createRequire(import.meta.url)('node:http');
 /** Max envelopes accepted from one peer in one sync. */
 export const MAX_ENVELOPES_PER_SYNC = 5_000;
 /** Max serialized bytes for a single envelope. */
@@ -389,7 +394,7 @@ export async function syncRoomFromPeer(basePath, peer, roomId, fetchImpl = fetch
  */
 export function serveFederation(basePath, opts = {}) {
     const host = opts.bindHost ?? '127.0.0.1';
-    const server = createServer((req, res) => {
+    const server = loadHttp().createServer((req, res) => {
         const send = (code, obj) => {
             const buf = Buffer.from(JSON.stringify(obj));
             res.writeHead(code, { 'content-type': 'application/json', 'content-length': buf.length });