Skip to content

claude-flow

v3.32.10 Feature

This release adds 2 notable features for engineering teams evaluating rollout.

✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

agentic-ai agentic-framework agentic-workflow agents ai-agents ai-assistant
+14 more
ai-coding ai-skills autonomous-agents claude-code codex harness mcp-server multi-agent multi-agent-systems npm skills swarm swarm-intelligence typescript

Summary

AI summary

Updates Commits, Test plan, and Investigated / not fixed across a mixed release.

Full changelog

Summary

Tracker-fire sweep from the 2026-07-24 → 2026-07-26 window. Ten tracked bugs plus one cold-start regression, landed as six commits on the branch (seventh test-coverage commit incoming). Rebased against origin/main at v3.32.9 — one issue's proposed fix (Codex #2774) was based on an incorrect diagnosis and reverted during rebase.

Regression check: apples-to-apples fresh-worktree tests — origin/main baseline 98 failed / 3199 total, this branch 91 failed / 3199 total — net -7 failures, zero new failures introduced.

Commits

  1. 38f4eaebf — main tracker sweep (9 fixes + promo seed)
  2. 8933c6c8cparser.ts:applyDefaults walks command/subcommand options (#2775 root cause)
  3. b2a383079memory_store MCP defaults upsert=true for CLI parity (#2775)
  4. 3abb25823 — CHANGELOG re-scope to 3.32.10
  5. 4c39225cahooks_post-task dual-writes routing decisions to JSON store (#2786 fix-2)
  6. 72e6cd705 — 3 parser regression tests for #2775 (52/52 → 55/55)

Fixed

  • Statusline promo row blank on new installs — cold-start local seed pool restored (ADR-311 had emptied it). Cognitum remote pool (funnel.ruv.io/v1/messages) remains authoritative via eligibleMessagesFromPools (remote wins by id); seed only renders during the ≤ 60s cold-start window before first successful remote fetch. Nine seed messages: 1 disclosure, 7 educational tips pointing at cognitum.one docs, 1 promotional. (v3/@claude-flow/cli/src/funnel/messages.ts)
  • #2777ruflo init no longer imports the entire ruvnet/ruflo repo (97 MB, 384 SKILL.md) into .agents/skills/ruflo/. Materializes single platform SKILL.md; detects bloated prior install and re-materializes.
  • #2781ruflo-adr adr-index no longer silently drops ADR data: status regex accepts - **Status**: proposed; single-line and wrapped relation lines parse fully; CLI_CORE=1 warns and unifies on @claude-flow/cli@latest. Dry-run over 531 ADRs captures 608 edges cleanly.
  • #2770 — Windows: browser-session MCP tools + two init execFileSync('npx', …) sites now set shell: process.platform === 'win32' so cmd.exe resolves npx.cmd. POSIX behavior unchanged.
  • #2782WorkerDaemon.saveState() + autopilot-state.saveState() + autopilot-state.appendLog() no longer race on shared .tmp filename. All three call writeFileAtomic (pid + timestamp + random-suffixed temp) with try/catch.
  • #2785ruflo hooks post-task accepts --task/-t and --store-results flags per CLAUDE.md-documented usage. Routing outcomes finally persist to the namespace hooks_metrics reads.
  • #2786 — AgentDB no longer silently fails to initialize under CLAUDE_FLOW_ENCRYPT_AT_REST=1. New getAgentDbPath() returns agentdb-memory.db in the same dir as memory.db; ControllerRegistry (native better-sqlite3) opens a distinct file from the sql.js CRUD writer's encrypted memory.db.
    • Follow-up #2786 fix-2: hooks_metrics "Pattern Learning" / "Agent Routing" counters were stuck at zero because the sync reader getIntelligenceStatsFromMemory() reads .claude-flow/memory/store.json while hooks_post-task writes only to AgentDB. Fix: dual-write — after AgentDB write, mirror routing decision into the JSON store with metadata.type: 'routing-decision'.
  • #2776 — Statusline security segment: STALE/IN_PROGRESS branches reachable via local overlay recomputing freshness on every render from .claude/security-scans/scan-*.json. Env: RUFLO_SCAN_STALE_HOURS (24), RUFLO_SCAN_PENDING_CAP_MIN (30). STALE renders dim gray.
  • #2775 — Memory store to existing key no longer dead-ends. Four layers of fix:
    1. bridgeStoreEntry uses INSERT ... ON CONFLICT with tombstone auto-resurrect; UNIQUE returns typed error instead of null (no more misleading demotion into #2735 guard).
    2. bridgeDeleteEntry runs wal_checkpoint(PASSIVE).
    3. parser.ts:applyDefaults walks command + subcommand options (root cause of --upsert default silently dropping — affected every subcommand-declared default, not just this one).
    4. memory_store MCP tool schema defaults upsert: true for CLI parity.

Investigated / not fixed

  • #2774 — Reporter's diagnosis "Codex MCP generator registers management CLI instead of stdio server" appears incorrect: ruflo mcp start IS a working stdio server per v3/@claude-flow/cli/src/commands/mcp.ts (MCP Server started on stdio). Upstream d20f1323b fix(codex): ship stable Windows-safe Ruflo integration cleanly centralized the Codex MCP config in mcp-config.ts using the same command shape. A proposed swap to claude-flow-mcp was reverted during rebase. Recommend closing unless a fresh repro against 3.32.9+ is produced.
  • #2786 fix-3bridgeRecordFeedback() calls learningSystem.recordFeedback() / .record() and reasoningBank.recordOutcome({taskId, verdict, …}) with signatures that don't exist on [email protected]'s actual API. Wrapped in silent catch. Genuine integration work — needs RL-session lifecycle management across separate CLI processes plus meaningful state/action/reward mapping. Deferred to a separate PR.

Test plan

  • [x] pnpm install at repo root
  • [x] cd v3/@claude-flow/security && npm run build → tsc + OAuth export verify pass
  • [x] cd v3/@claude-flow/cli && npm run build → tsc exit 0
  • [x] cd v3/@claude-flow/codex && npm run build → tsc exit 0
  • [x] CLI test suite on fresh worktree: 91 failed / 3031 passed / 77 skipped / 3199 total — vs origin/main baseline 98 failed / same total. Net -7 failures; sampled failing tests are WASM/sharp/env issues in files this PR does not touch.
  • [x] Parser tests: 55/55 pass (52 pre-existing + 3 new #2775 regression tests)
  • [ ] Windows CI: exercise browser-session MCP tools and init execFileSync npx paths to confirm the shell: process.platform === 'win32' fix.
  • [ ] Fresh install smoke: ruflo init in an empty dir, check .agents/skills/ruflo/ is a single ~2KB SKILL.md (not 97MB).
  • [ ] Statusline cold-start: on a clean ~/.ruflo/, first ruflo hooks statusline render shows a promo row from the local seed pool.
  • [ ] Encrypt-at-rest: CLAUDE_FLOW_ENCRYPT_AT_REST=1 ruflo memory store … succeeds and agentdb-memory.db appears next to memory.db.
  • [ ] hooks_metrics after several hooks_post-task --store-results calls: "Agent Routing" decision count is non-zero.

Closes: #2770 #2776 #2777 #2781 #2782 #2785 #2786 (main + fix-2)
Refs: #2775 (main fix inline + parser root cause + MCP parity)
Investigation: #2774 (proposed fix reverted; diagnosis appears incorrect against current main)
Deferred: #2786 fix-3 (agentdb LearningSystem API skew — architectural)

🤖 Generated with RuFlo

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Track claude-flow

Get notified when new releases ship.

Sign up free

About claude-flow

Deploy multi-agent swarms with coordinated workflows.

All releases →

Related context

Beta — feedback welcome: [email protected]