@sylphx/flow 1.4.6 → 1.4.8

package/assets/rules/workspace.md

@@ -0,0 +1,485 @@
+---
+name: Workspace Documentation
+description: .sylphx/ shared workspace - auto-create, maintain SSOT, verify on every read
+---
+
+# WORKSPACE DOCUMENTATION
+
+## On First Task
+
+**Check:** `.sylphx/` exists?
+
+**No → Create structure:**
+```bash
+mkdir -p .sylphx/decisions
+```
+
+Create files with templates below. Populate with project-specific content.
+
+**Yes → Verify:**
+- Read all files
+- Check accuracy vs actual code
+- Update or delete outdated sections
+
+---
+
+## Structure & Templates
+
+### .sylphx/context.md
+
+**Create when:** First task, or when missing
+**Update when:** Project scope/purpose/constraints change
+
+```markdown
+# Project Context
+
+## What
+[1-2 sentence description of what this project is]
+
+## Why
+[Problem being solved, user need addressed]
+
+## Who
+[Target users, primary use cases]
+
+## Status
+[Development phase: Alpha/Beta/Stable, current version]
+
+## Key Constraints
+- [Non-negotiable requirement 1]
+- [Non-negotiable requirement 2]
+- [Critical limitation or boundary]
+
+## Source of Truth References
+<!-- VERIFY: These files exist -->
+- Tech stack: `package.json`
+- Configuration: [list config files]
+- Build/Scripts: `package.json` scripts
+```
+
+**Verify:** Referenced files exist. If not, update or remove reference.
+
+---
+
+### .sylphx/architecture.md
+
+**Create when:** First task, or when missing
+**Update when:** Architecture changes, patterns adopted, major refactoring
+
+```markdown
+# Architecture
+
+## System Overview
+[1-2 paragraph high-level description]
+
+## Key Components
+<!-- VERIFY: Paths exist -->
+- **Component A** (`src/path/`): [Purpose, responsibility]
+- **Component B** (`src/path/`): [Purpose, responsibility]
+
+## Design Patterns
+
+### Pattern: [Name]
+**Why chosen:** [Rationale - problem it solves]
+**Where used:** `src/path/to/implementation.ts`
+**Trade-off:** [What gained vs what lost]
+
+## Data Flow
+[Macro-level: input → processing → output]
+See `src/[entry-point].ts` for implementation.
+
+## Boundaries
+**In scope:** [What this project does]
+**Out of scope:** [What it explicitly doesn't do]
+```
+
+**Verify:** All paths exist. Patterns still used. Trade-offs still accurate.
+
+---
+
+### .sylphx/glossary.md
+
+**Create when:** First task, or when missing
+**Update when:** New project-specific term introduced
+
+```markdown
+# Glossary
+
+## [Term]
+**Definition:** [Clear, concise definition]
+**Usage:** `src/path/where/used.ts`
+**Context:** [When/why this term matters]
+
+---
+
+[Only project-specific terms. No general programming concepts.]
+```
+
+**Verify:** Terms still used. Usage references exist.
+
+---
+
+### .sylphx/decisions/README.md
+
+**Create when:** First ADR created
+**Update when:** New ADR added
+
+```markdown
+# Architecture Decision Records
+
+## Active Decisions
+- [ADR-001: Title](./001-title.md) ✅ Accepted
+- [ADR-002: Title](./002-title.md) ✅ Accepted
+
+## Superseded
+- [ADR-XXX: Old Title](./xxx-old.md) 🔄 Superseded by ADR-YYY
+
+## Status Legend
+- ✅ Accepted - Currently in effect
+- ⏸️ Proposed - Under consideration
+- ❌ Rejected - Not adopted
+- 🔄 Superseded - Replaced by newer ADR
+```
+
+---
+
+### .sylphx/decisions/NNN-title.md
+
+**Create when:** Making architectural decision
+**Update when:** Decision status changes or is superseded
+
+```markdown
+# NNN. [Title - Verb + Object, e.g., "Use Bun as Package Manager"]
+
+**Status:** ✅ Accepted
+**Date:** YYYY-MM-DD
+**Deciders:** [Who made decision, or "Project maintainers"]
+
+## Context
+[Situation/problem requiring a decision. 1-2 sentences.]
+
+## Decision
+[What was decided. 1 sentence.]
+
+## Rationale
+[Why this decision over alternatives. Key benefits. 2-3 bullet points.]
+
+## Consequences
+**Positive:**
+- [Benefit 1]
+- [Benefit 2]
+
+**Negative:**
+- [Drawback 1]
+- [Drawback 2]
+
+## References
+<!-- VERIFY: Links exist -->
+- Implementation: `src/path/to/code.ts`
+- Related PR: #123 (if applicable)
+- Supersedes: ADR-XXX (if applicable)
+```
+
+**Keep <200 words total.**
+
+---
+
+## SSOT Discipline
+
+**Never duplicate. Always reference.**
+
+### ❌ Bad (Duplication - Will Drift)
+
+```markdown
+Dependencies:
+- react 19.2.0
+- zod 4.1.12
+
+Linting rules:
+- no-unused-vars
+- prefer-const
+```
+
+### ✅ Good (Reference - SSOT Maintained)
+
+```markdown
+<!-- VERIFY: package.json exists -->
+Dependencies: See `package.json`
+
+<!-- VERIFY: biome.json exists -->
+Linting: Biome (config in `biome.json`)
+
+## Why Biome
+- Decision: ADR-003
+- Benefit: Single tool for format + lint
+- Trade-off: Smaller ecosystem than ESLint
+```
+
+**Format for references:**
+```markdown
+<!-- VERIFY: path/to/file.ts -->
+[Description]. See `path/to/file.ts`.
+```
+
+Verification marker reminds: when file changes, check if doc needs update.
+
+---
+
+## Maintenance Triggers
+
+### On Every Task Start
+
+```
+1. Check .sylphx/ exists
+   - No → Create with templates
+   - Yes → Continue to verify
+
+2. Read all .sylphx/ files
+
+3. Verify accuracy:
+   - Check <!-- VERIFY: --> markers
+   - Confirm files exist
+   - Check if still accurate vs code
+
+4. Update or delete:
+   - Wrong → Fix immediately
+   - Outdated → Update or delete
+   - Missing context → Add
+
+5. Note gaps for later update
+```
+
+### During Task Execution
+
+**Triggers to update:**
+
+- **New understanding** → Update context.md or architecture.md
+- **Architectural decision made** → Create ADR in decisions/
+- **New project-specific term** → Add to glossary.md
+- **Pattern adopted** → Document in architecture.md with WHY
+- **Constraint discovered** → Add to context.md
+- **Found outdated info** → Delete or update immediately
+
+### Before Commit
+
+```
+1. Updated understanding? → Update .sylphx/
+2. Made architectural change? → Create/update ADR
+3. Deprecated approach? → Mark superseded or delete
+4. Verify: No contradictions between .sylphx/ and code
+5. Verify: All <!-- VERIFY: --> markers still valid
+```
+
+---
+
+## Content Rules
+
+### ✅ Include (Macro-Level WHY)
+
+- Project purpose and context
+- Architectural decisions (WHY chosen)
+- System boundaries (in/out of scope)
+- Key patterns (WHY used, trade-offs)
+- Project-specific terminology
+- Non-obvious constraints
+
+### ❌ Exclude (Belongs Elsewhere)
+
+- API documentation → JSDoc in code
+- Implementation details → Code comments
+- Configuration values → Config files
+- Dependency versions → package.json
+- Code examples → Actual code or tests
+- How-to guides → Code comments
+- Step-by-step processes → Code itself
+
+**Principle:** If it's in code or config, don't duplicate it here.
+
+---
+
+## Red Flags (Delete Immediately)
+
+Scan for these on every read:
+
+- ❌ "We plan to..." / "In the future..." (speculation)
+- ❌ "Currently using..." (implies might change - use present tense or delete)
+- ❌ Contradicts actual code
+- ❌ References non-existent files
+- ❌ Duplicates package.json / config
+- ❌ Explains HOW instead of WHY
+- ❌ Generic advice (not project-specific)
+
+**When found:** Delete entire section immediately.
+
+---
+
+## Cleanup Protocol
+
+**Monthly or after major changes:**
+
+```bash
+# 1. Check all referenced files exist
+cd .sylphx
+grep -r "src/" . | grep -o 'src/[^`)]*' | sort -u > /tmp/refs.txt
+# Verify each file in refs.txt exists
+
+# 2. Check package.json references
+grep -r "package.json" .
+# Verify info isn't duplicated
+
+# 3. Check verification markers
+grep -r "<!-- VERIFY:" .
+# Check each marked file exists and content accurate
+
+# 4. Read all files
+# Delete outdated sections
+# Update inaccurate content
+# Remove speculation
+```
+
+---
+
+## Decision Flow: Create ADR?
+
+**Create ADR when:**
+- Choosing between 2+ significant alternatives
+- Decision has long-term impact
+- Future developers will ask "why did they do this?"
+- Non-obvious trade-offs involved
+
+**Don't create ADR for:**
+- Obvious choices (use standard tool)
+- Temporary decisions (will change soon)
+- Implementation details (belongs in code comments)
+- Trivial choices (naming, formatting)
+
+**Quick test:** Will this decision matter in 6 months? Yes → ADR. No → Skip.
+
+---
+
+## Verification Commands
+
+**Check links valid:**
+```bash
+cd .sylphx
+# Extract all file references
+grep -roh '`[^`]*\.[a-z]*`' . | tr -d '`' | sort -u | while read f; do
+  [ -f "../$f" ] || echo "MISSING: $f"
+done
+```
+
+**Check for duplication:**
+```bash
+# If package.json mentioned without "See package.json"
+grep -r "dependencies" .sylphx/ | grep -v "See \`package.json\`"
+# Should be empty or references only
+```
+
+---
+
+## Examples
+
+### Good context.md (Real Project)
+
+```markdown
+# Project Context
+
+## What
+AI-powered CLI for autonomous development workflows with agent orchestration.
+
+## Why
+Enable developers to delegate complex multi-step tasks to AI that can plan, execute, verify autonomously while maintaining quality.
+
+## Who
+Developers using Claude/AI for coding assistance.
+
+## Status
+Active development - v1.2.0
+Focus: Agent prompt optimization
+
+## Key Constraints
+- No breaking changes without major version
+- Research mandatory before implementation
+- All modules need .test.ts and .bench.ts
+- Clean commits only (no TODOs, debug code)
+
+## Source of Truth
+<!-- VERIFY: packages/flow/package.json -->
+- Dependencies: `packages/flow/package.json`
+- Build: `package.json` scripts (root + packages/flow)
+- TypeScript: `packages/flow/tsconfig.json`
+```
+
+### Good architecture.md
+
+```markdown
+# Architecture
+
+## System Overview
+CLI loads agent prompts from markdown, composes with rules/output-styles, orchestrates multi-agent workflows.
+
+## Key Components
+<!-- VERIFY: Paths exist -->
+- **Agent Loader** (`src/core/agent-loader.ts`): Parses markdown prompts
+- **Agent Manager** (`src/core/agent-manager.ts`): Orchestration
+
+## Design Patterns
+
+### Pattern: Markdown-as-Config
+**Why:** Human-readable, version-controlled, easy iteration
+**Where:** `assets/**/*.md` with frontmatter
+**Trade-off:** Parsing overhead vs flexibility (chose flexibility)
+
+### Pattern: Minimal Effective Prompting
+**Why:** Trust LLM, reduce tokens 40%, increase clarity
+**Where:** All prompts (v1.2.0 refactor)
+**Trade-off:** Less explicit teaching vs more effective triggering
+**Decision:** ADR-002
+```
+
+### Good ADR
+
+```markdown
+# 002. Minimal Effective Prompt Philosophy
+
+**Status:** ✅ Accepted
+**Date:** 2024-11-15
+
+## Context
+Agent prompts were verbose with step-by-step teaching, reducing effectiveness and increasing cost.
+
+## Decision
+Adopt MEP: Trust LLM, WHAT+WHEN not HOW+WHY, mixed formats, condensed.
+
+## Rationale
+- 40% token reduction
+- Better LLM performance (less noise)
+- Easier maintenance
+
+## Consequences
+**Positive:** Lower cost, better results, cleaner prompts
+**Negative:** Less explicit for human readers
+
+## References
+<!-- VERIFY: commit exists -->
+- Implementation: All `assets/**/*.md` files
+- Refactor: commit c7795c0f
+```
+
+---
+
+## Summary
+
+**Agent behavior:**
+1. **First task:** Check .sylphx/ exists → Create if missing → Populate with templates
+2. **Every task start:** Read .sylphx/ → Verify accuracy → Update/delete as needed
+3. **During work:** New understanding → Update immediately
+4. **Before commit:** Verify .sylphx/ matches reality → No contradictions
+
+**Content:**
+- **Include:** WHY (context, decisions, rationale)
+- **Exclude:** HOW (implementation → code)
+- **Reference:** Link to SSOT, never duplicate
+- **Maintain:** Verify on read, update on learn, delete when wrong
+
+**Prime Directive: Outdated docs worse than no docs. When in doubt, delete.**

package/assets/slash-commands/cleanup.md

@@ -0,0 +1,60 @@
+---
+name: cleanup
+description: Clean technical debt, refactor, optimize, and simplify codebase
+agent: coder
+---
+
+# Cleanup & Refactor
+
+Scan codebase for technical debt and code smells. Clean, refactor, optimize.
+
+## Scope
+
+**Code Smells:**
+- Functions >20 lines → extract
+- Duplication (3+ occurrences) → DRY
+- Complexity (>3 nesting levels) → flatten
+- Unused code/imports/variables → remove
+- Commented code → delete
+- Magic numbers → named constants
+- Poor naming → clarify
+
+**Technical Debt:**
+- TODOs/FIXMEs → implement or delete
+- Deprecated APIs → upgrade
+- Outdated patterns → modernize
+- Performance bottlenecks → optimize
+- Memory leaks → fix
+- Lint warnings → resolve
+
+**Optimization:**
+- Algorithm complexity → reduce
+- N+1 queries → batch
+- Unnecessary re-renders → memoize
+- Large bundles → code split
+- Unused dependencies → remove
+
+## Execution
+
+1. **Scan** entire codebase systematically
+2. **Prioritize** by impact (critical → major → minor)
+3. **Clean** incrementally with atomic commits
+4. **Test** after every change
+5. **Report** what was cleaned and impact
+
+## Commit Strategy
+
+One commit per logical cleanup:
+- `refactor(auth): extract validateToken function`
+- `perf(db): batch user queries to fix N+1`
+- `chore: remove unused imports and commented code`
+
+## Exit Criteria
+
+- [ ] No code smells remain
+- [ ] All tests pass
+- [ ] Lint clean
+- [ ] Performance improved (measure before/after)
+- [ ] Technical debt reduced measurably
+
+Report: Lines removed, complexity reduced, performance gains.

package/assets/slash-commands/improve.md

@@ -0,0 +1,154 @@
+---
+name: improve
+description: Proactively discover and implement improvements
+agent: coder
+---
+
+# Proactive Improvement
+
+Analyze project comprehensively. Discover improvement opportunities. Prioritize and execute.
+
+## Discovery Areas
+
+**Code Quality:**
+- Code complexity hotspots
+- Duplication patterns
+- Test coverage gaps
+- Error handling weaknesses
+- Type safety improvements
+- Naming inconsistencies
+
+**Architecture:**
+- Tight coupling
+- Missing abstractions
+- Scalability bottlenecks
+- State management issues
+- API design improvements
+- Module organization
+
+**Performance:**
+- Slow algorithms (profiling data)
+- Database query inefficiencies
+- Bundle size optimization
+- Memory usage patterns
+- Network request optimization
+- Caching opportunities
+
+**Security:**
+- Vulnerability scan (dependencies)
+- Input validation gaps
+- Authentication/authorization weaknesses
+- Sensitive data exposure
+- OWASP top 10 check
+- Security best practices
+
+**Developer Experience:**
+- Development setup complexity
+- Build time optimization
+- Testing speed
+- Debugging capabilities
+- Error messages clarity
+- Documentation gaps
+
+**Maintenance:**
+- Outdated dependencies
+- Deprecated API usage
+- Missing tests
+- Incomplete documentation
+- Configuration complexity
+- Technical debt accumulation
+
+**Features:**
+- Missing functionality (user feedback)
+- Integration opportunities
+- Automation potential
+- Monitoring/observability
+- Error recovery
+- User experience improvements
+
+## Analysis Process
+
+1. **Scan** codebase comprehensively
+2. **Profile** performance bottlenecks
+3. **Check** security vulnerabilities
+4. **Review** dependencies for updates
+5. **Analyze** test coverage
+6. **Assess** documentation completeness
+7. **Evaluate** architectural patterns
+8. **Identify** missing features
+
+## Prioritization Matrix
+
+**Impact vs Effort:**
+- **High Impact + Low Effort** → Do first
+- **High Impact + High Effort** → Plan carefully
+- **Low Impact + Low Effort** → Quick wins
+- **Low Impact + High Effort** → Skip or defer
+
+**Categories:**
+- **Critical:** Security, data loss, crashes
+- **High:** Performance degradation, poor UX
+- **Medium:** Code quality, maintainability
+- **Low:** Nice-to-haves, polish
+
+## Execution Plan
+
+For each improvement:
+1. **Describe** what and why
+2. **Estimate** effort (small/medium/large)
+3. **Assess** impact (low/medium/high)
+4. **Prioritize** using matrix
+5. **Create** implementation plan
+
+Then execute top priorities:
+- Start with high-impact, low-effort
+- Make atomic commits
+- Test thoroughly
+- Document changes
+
+## Output Format
+
+```markdown
+## Improvement Opportunities
+
+### Critical (Do Now)
+1. [Security] Update vulnerable dependency X (CVE-XXXX)
+   - Impact: High (prevents exploit)
+   - Effort: Low (5 min)
+   - Action: `npm update X`
+
+### High Priority
+1. [Performance] Optimize user query (N+1 problem)
+   - Impact: High (10x speedup)
+   - Effort: Medium (2 hrs)
+   - Action: Batch queries with JOIN
+
+### Quick Wins
+1. [DX] Add pre-commit hooks
+   - Impact: Medium (catch errors early)
+   - Effort: Low (30 min)
+   - Action: Setup husky + lint-staged
+
+### Planned Improvements
+1. [Feature] Add caching layer
+   - Impact: High (reduce server load)
+   - Effort: High (1 week)
+   - Action: Design cache strategy, implement Redis
+
+## Execution Plan
+1. Fix security vulnerability (now)
+2. Optimize query performance (today)
+3. Add pre-commit hooks (today)
+4. Plan caching implementation (next sprint)
+```
+
+## Exit Criteria
+
+- [ ] Comprehensive analysis completed
+- [ ] Opportunities prioritized
+- [ ] High-priority improvements implemented
+- [ ] Tests pass
+- [ ] Impact measured
+- [ ] Documentation updated
+
+Report: Improvements discovered, implemented, impact metrics.