devflow-research

Devflow

The problem with AI-assisted development

Claude Code is powerful. But every session starts from scratch. Context evaporates between conversations. Code reviews are single-pass and shallow. Quality depends entirely on what you remember to ask for.

Devflow fixes this. Install once, forget about it. Your code gets better automatically.

It detects when you paste a structured plan so it can auto-execute — no classification overhead for normal prompts. Complex tasks get an advanced TDD and EDD harness with quality gates at every step.

See it work

you: add rate limiting to the /api/upload endpoint
     (or use /implement for the full agent pipeline)

         → Created branch feat/42-rate-limit-upload
         → Exploring codebase... Planning... Coding...
         → Validator: build ✓ typecheck ✓ lint ✓ tests ✓
         → Simplifier: cleaned up 3 files
         → Scrutinizer: 9-pillar quality check passed
         → Evaluator: implementation matches request ✓
         → Tester: 5/5 QA scenarios passed ✓

/code-review     → 18 reviewers examine your changes in parallel
/resolve         → all issues validated and fixed automatically
/bug-analysis    → proactive bug finding before review

What you get

Ambient intelligence. Devflow detects workflow intent — start a prompt with implement, explore, research, debug, or plan, or paste a structured plan, and the matching workflow runs automatically. Normal prompts get zero overhead, no classification step. No slash commands needed. Init and forget.

Memory that persists. Session context survives restarts, /clear, and context compaction. Your AI picks up exactly where it left off. Architectural decisions and known pitfalls accumulate in .devflow/decisions/ and inform every future session. No manual bookkeeping.

Decisions accumulate automatically. A background agent detects architectural decisions and known pitfalls from your session dialogs and writes them to .devflow/decisions/decisions.md and .devflow/decisions/pitfalls.md — informing every future review and implementation session without any manual bookkeeping.

18 parallel code reviewers. Security, architecture, performance, complexity, consistency, regression, testing, and more. Each produces findings with severity, confidence scoring, and concrete fixes. Conditional reviewers activate when relevant (TypeScript for .ts files, database for schema changes). Every finding gets validated and resolved automatically.

45 skills. 41 are grounded in expert material — backed by peer-reviewed papers, canonical books, and industry standards: security (OWASP, Shostack), architecture (Parnas, Evans, Fowler), performance (Brendan Gregg), testing (Beck, Meszaros), design (Wlaschin, Hickey), 200+ sources total. The remaining 4 are procedural Dream maintenance skills.

Skill shadowing. Override any built-in skill with your own version. Drop a file into ~/.devflow/skills/{name}/ and the installer uses yours instead of the default — same activation, your rules.

Always-on rules. 12 ultra-condensed engineering principles (~10 lines each) load on every prompt — security, quality, and language-specific guidance (TypeScript, React, Go, Python, Java, Rust). Rules install from your selected plugins only, so a Go project won't get React rules. Override any rule via ~/.devflow/rules/{name}.md.

Full lifecycle. /plan takes a feature idea through codebase exploration, gap analysis, design review, and outputs a plan document ready for /implement. /implement accepts that plan document (or an issue or task description directly) and drives it through coding, validation, and refinement to a PR. /debug investigates bugs with competing hypotheses in parallel. /self-review runs Simplifier + Scrutinizer quality passes.

Everything is composable. 22 plugins (12 core + 9 language/ecosystem + 1 optional workflow). Install only what you need.

HUD. A persistent status line updates on every prompt — project, branch, diff stats, context usage, model, cost with weekly/monthly totals, quota reset timers, and configuration counts at a glance.