This repository was archived by the owner on Apr 11, 2026. It is now read-only.
Releases: z23cc/flow-code
Releases · z23cc/flow-code
v0.1.51
Pipeline Hygiene
Added
- Epic review step (Step 4.5) in Close phase — verifies spec compliance before shipping
- Shared review protocol references in plan-review, impl-review, and epic-review skills
Changed
CLAUDE.mddocumentsskills/vscodex/skills/directory responsibilities and priority rules- Work phase audit warnings replaced with clean safety invariants (verified fixed)
Cleaned
- Archived 13 stale/test epics from .flow/ state
Full Changelog: v0.1.50...v0.1.51
v0.1.50
RP Orchestrate Pattern Adoption
Adopts 4 key patterns from RepoPrompt's Orchestrate workflow:
Parallel Explore Agents (Brainstorm)
- 3 concurrent explore agents (code patterns, gaps/debt, dependency impact) run before
context_builder - Multi-dimensional codebase analysis instead of single-perspective brainstorm
Plan-as-Contract (Worker Prompt)
flowctl worker-promptnow outputs aPLAN CONTRACTpreamble- Workers must read the epic spec first and implement only their assigned task
- Reduces worker drift from plan intent
Concurrency Awareness (Worker Prompt)
CONCURRENT_WORKERSblock lists other active tasks and their locked files- Workers avoid modifying files owned by concurrent workers
- Conflict escalation: workers output
STATUS: needs_file_accesswhen needing a locked file
Worker Phase 2 Restructure
- Phase 2 now has 3 sub-phases: plan-as-contract (2.0), concurrency awareness (2.1), read specs (2.2)
- Mandatory plan spec citation before implementation begins
Full Changelog: v0.1.49...v0.1.50
v0.1.49
What's Changed
- RP agent_run pipeline — Worker spawning via RepoPrompt agent sessions
- New commands —
/flow-code:adr(architecture decisions) and/flow-code:spec(requirements capture) - Shared RP/MCP orchestration guide — Canonical guidance for RepoPrompt tool usage
- Context engineering skill — Token management and context window optimization
- Expanded skills — brainstorm, export-context, rp-explorer, run, and more
- flowctl improvements — Core+CLI enhancements across the Rust codebase
- CI/release checks — Version parity preflight script
Full Changelog: v0.1.48...v0.1.49
v0.1.48 — RP agent_run pipeline + 8 optimizations
What's New
RP agent_run replaces Claude Agent/Teams
- Workers spawned via
agent_run(engineer/pair/explore)— multi-model support - Workers self-create worktrees (coordinator only does
start+merge) steerreplacesSendMessagefor cross-worker coordination- Continuous streaming flow replaces wave-based execution
- Codex sandbox approval auto-handled via
respond(accept_for_session)
8 Pipeline Optimizations
| # | Optimization | Effect |
|---|---|---|
| 1 | flowctl startup |
6 CLI calls → 1 |
| 2 | Brainstorm+Plan research dedup | context_builder + oracle_send reuse across phases |
| 3 | Speculative execution | Plan Review + Work Wave 1 in parallel |
| 4 | worker-phase next --all |
24 CLI calls → 1 (batch all phases) |
| 5 | worker-prompt --inline-skills |
Core rules embedded in prompt (no Phase 2 file reads) |
| 6 | checklist check-all |
8 CLI calls → 1 (batch verification) |
| 7 | Review chat_id reuse |
Impl Review uses oracle_send to continue Plan Review context |
| 8 | Worker Phase 10 guard dedup | Skip guard if no commits since Phase 6 |
New flowctl Commands
flowctl startup --json— combined detect + interrupted + review-backend + branch + epicsflowctl worker-phase next --all --json— return all remaining phases at onceflowctl worker-prompt --inline-skills— embed core skill rules in bootstrap promptflowctl checklist check-all --items "a,b,c"— batch check multiple items
Files Changed
18 files, +776/-659 lines
Full Changelog: v0.1.32...v0.1.48
v0.1.32
What's Changed
- feat: Investigation targets in task specs by @z23cc in #11
- feat: Codex cross-platform support (Phase 1) by @z23cc in #12
- feat(flowctl): declarative Changes pattern, bidirectional dep index, --dry-run by @z23cc in #13
- Borrow agent-skills patterns: anti-rationalization, red flags, verification, references by @z23cc in #15
- feat: add engineering discipline skills inspired by agent-skills by @z23cc in #17
- refactor: flatten workflow phase/step numbering to sequential integers by @z23cc in #18
- feat(flowctl): event-sourced pipeline-first architecture by @z23cc in #19
- chore: consolidate plugin — remove deprecated skills, codex dupes, fix refs by @z23cc in #20
- chore: remove deprecated skill/command remnants by @z23cc in #21
- chore: remove all deprecated skill references by @z23cc in #22
- feat(flowctl): remove libSQL/fastembed — pure file-based state by @z23cc in #23
- refactor(flowctl): merge 4 crates into 2 (core + cli) by @z23cc in #24
- docs: remove stale libSQL/async references by @z23cc in #25
Full Changelog: v0.1.29...v0.1.32
v0.1.29
Full Changelog: v0.1.26...v0.1.29
v0.1.26
What's Changed
- Flowctl comprehensive optimization and hardening by @z23cc in #1
- Enforce Teams mode as default parallel execution by @z23cc in #2
- V2 reliability and performance hardening by @z23cc in #3
- Structured review findings with auto gap registration by @z23cc in #4
- feat(flowctl-rust): platform-grade Rust rewrite — 6 crates, 26k lines, 224 tests by @z23cc in #5
- feat: 智能编排引擎升级 — 服务层统一 + 自适应调度 + DAG可视化 + 自省 + 跨模型审查 by @z23cc in #6
- chore(frontend): switch to bun (remove pnpm/node refs) by @z23cc in #7
- feat(web): reposition UI as viewer + light CRUD by @z23cc in #8
- feat(flowctl): migrate rusqlite → libSQL async + native vectors by @z23cc in #9
- chore(flowctl): rename flowctl-db-lsql → flowctl-db by @z23cc in #10
New Contributors
Full Changelog: https://github.com/z23cc/flow-code/commits/v0.1.26