IDD + BDD + TDD for AI-generated code. Quality gates, blast radius analysis, and intent-driven development. The AI discipline layer for Claude Code (Cursor: stable v5.1 tier).
/plugin marketplace add galando/temper /plugin install temper
Commands available in any project.
bash <(curl -fsSL https://raw.githubusercontent.com/galando/temper/main/scripts/install-cursor.sh)
Stable v5.1 feature set. New capabilities ship Claude Code-first.
Temper runs the complete software development lifecycle — plan, design, build, review, check — with stage gates, feedback loops, and observability built in.
Review finds issues? Loop back to Build. Check fails tests? Targeted fix. Bad plan? Loop back to Plan. Circuit breakers: max 2 loops, same issue twice = stop.
System architecture, API contracts, database schema, integration points, and decision logs. Auto-skipped for simple features, enforced for complex ones.
Every scenario runs against your test runner. Real pass/fail, real output. Not opinions — proof. Evidence labels: [PROVEN], [HEURISTIC], [SEMANTIC].
Learns from every review. Detects recurring patterns, suggests new pack rules, and suppresses noise. Your reviews get sharper over time.
Each stage writes structured context for downstream stages. Review knows what Build struggled with, Check knows what Review already fixed. No agent amnesia.
Walk through plans and designs section by section — intent, scenarios, architecture, blast radius, tasks. Skip to build from any point.
Optional MCP servers upgrade analysis: code-review-graph for blast radius and call chains, semgrep for SAST scanning. Progressive enhancement.
Temper for Cursor at the v5.1 feature set. 22 rules — packs, skills, capabilities. New capabilities ship Claude Code-first. Platform strategy →
Orchestrates Alibaba's open-code-review for LLM-powered defect detection. Cross-validated findings labeled [OCR+TEMPER] with boosted confidence.
Architecture arguments are fine. Results are better. Here's what Temper catches that vanilla AI coding misses.
AI built password reset. All tests pass. Scenario coverage gate caught the gap: no test for rate limiting. Build wrote the test → failed → implemented rate limiting → passed.
AI planned 3 classes for a single validation rule. File-to-scenario traceability flagged it: one scenario, one function needed. Three files became zero.
AI built the feature correctly — plus an admin-only endpoint nobody asked for. Untraced file flagged and removed before commit.
Most AI tools only answer "does the code work?" Temper answers all three questions that matter.
Did we solve the problem? Success criteria with structured validation types. Not just "tests pass" — the feature actually addresses the need.
Does it do the right things? Scenarios derived BEFORE architecture. Every file traces to a behavior. No over-engineering.
Does the code work? Tests written from scenarios. RED → GREEN → REFACTOR. Scenario coverage gate ensures nothing is missed.
All three layers live in a single markdown artifact. Problem, scenarios, and coverage checklist in one place.
Temper doesn't just generate code — it plans, validates, tests, and learns. Every step is verified before moving forward.
Understands impact before coding. Maps affected files, dependencies, and risk areas.
Auto-generates flowcharts, sequence diagrams, and state machines. See the plan visually before building.
Walk through the plan section by section — intent, scenarios, architecture, blast radius, tasks — with Q&A at each step.
Tests must pass. Coverage must be met. The AI cannot proceed until verified.
Four steps from idea to production-ready code. Temper automates quality — you focus on shipping.
/temperOne command starts it all. Temper analyzes blast radius, generates mermaid architecture diagrams, creates BDD scenarios, and presents an interactive plan. Walk through step by step or approve and build.
Temper implements automatically with TDD gates enforced at every step. Tests written first, code follows. Stage gate asks to continue.
Temper reviews automatically against enabled packs. Findings scored by confidence. Stage gate confirms all checks passed.
Temper commits your changes with a proper message. Branch ready to push. Use /temper:status anytime to track quality metrics.
Temper gets smarter with every review. It detects recurring patterns, suggests new rules, and suppresses false positives — so your reviews focus on what actually matters.
Identifies recurring issues and anti-patterns in your code.
Proposes new rules based on what it learns from your reviews.
Suppresses false positives as it learns your codebase patterns.
Rules that enforce best practices. Each pack has BLOCK, WARN, and SUGGEST severity levels.
Add a rules.md file and Temper will automatically discover it. Three-tier resolution: project > global > built-in. Link packs to plugins or skills for automatic context injection.
.claude/packs/my-company/rules.md
# My Company Standards
## BLOCK
- All API responses use DTOs
- No raw SQL queries
## WARN
- Constructor injection only
- Max method length: 20 lines
Type /temper:pack in any project
Temper scans your codebase and asks about patterns it finds
Temper generates a custom rules.md based on your answers
One command runs the full SDLC: plan → design → build → review → check with stage gates, feedback loops, and observability.
/temper "add login feature" # Full SDLC in one command
Runs plan, build, review, and check with interactive stage gates. Each stage asks for confirmation before proceeding.
For granular control, use individual commands.
Plan with blast radius, mermaid diagrams, and interactive walkthrough
System design for complex features (auto-skipped for simple)
Build with TDD + quality gates
Review with confidence scoring
Stack validation (auto-detect)
Root cause analysis + fix
Manage quality packs
Quality metrics dashboard
Unified: plan → design → build → review → check
Auto-detection with zero configuration.
/plugin marketplace add galando/temper /plugin install temper
Commands are now available in any project. For optional MCP-powered analysis, see Recommended Setup.
bash <(curl -fsSL https://raw.githubusercontent.com/galando/temper/main/scripts/install-cursor.sh)
Downloads 22 rules and 9 commands — packs, skills, capabilities, and reference docs. Stable v5.1 feature set. No Python or git required.