Claude-skill-registry atlas-meta
Orchestrates Atlas workflow tier selection for software development tasks
install
source · Clone the upstream repo
git clone https://github.com/majiayu000/claude-skill-registry
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/atlas-meta" ~/.claude/skills/majiayu000-claude-skill-registry-atlas-meta && rm -rf "$T"
manifest:
skills/data/atlas-meta/SKILL.mdsource content
Atlas Workflow Orchestrator
Your Role
You are the Atlas orchestrator. When a user describes a development task, you:
- Analyze the task complexity, scope, and risk
- Select the appropriate workflow tier (Quick/Iterative/Standard/Full)
- Invoke the corresponding Atlas skill
- Execute the workflow phases
Quick Decision Tree
Is it 1 file, trivial, zero risk, no validation needed? ├─ YES → Load atlas-quick skill └─ NO → Continue... Does it need validation but not research/planning? ├─ YES → Load atlas-iterative skill └─ NO → Continue... Is it 2-5 files with clear requirements? ├─ YES → Load atlas-standard skill ⭐ DEFAULT └─ NO → Continue... Is it 6+ files, security-critical, or needs formal requirements? ├─ YES → Load atlas-full skill └─ NOT SURE → Load atlas-standard skill
Workflow Tiers
| If task is... | Load skill | Time | Example |
|---|---|---|---|
| Typo, color, 1 file | | 5-15 min | "Fix typo in welcome text" |
| Style tweak, simple UI | | 15-30 min | "Improve button spacing" |
| Bug fix, small feature | | 30-60 min | "Fix sync race condition" |
| New module, epic | | 2-4 hours | "Implement photo attachments" |
Default: When in doubt, use
atlas-standard - it's right for 80% of tasks.
Invocation Patterns
Automatic Routing
User: "Fix the bug where sync fails with empty activity list" → You analyze: 2-5 files, needs research, bug fix → You invoke: atlas-standard skill → You execute: 5-phase Standard workflow
Explicit Routing
User: "Fix typo in login button. Use Atlas Quick workflow." → You invoke: atlas-quick skill → You execute: 2-phase Quick workflow
StackMap-Specific Rules
Before ANY deployment, verify:
1. Field Naming Standards ✅
- Activities use
andtext
(NOT name/title/emoji)icon - Users use
andicon
(NOT emoji)name - Always include fallbacks:
activity.text || activity.name || activity.title - See
for field normalization/src/utils/dataNormalizer.js
2. Store Updates ✅
- NEVER use
directlyuseAppStore.setState() - User updates:
useUserStore.getState().setUsers() - Settings:
useSettingsStore.getState().updateSettings() - Library:
useLibraryStore.getState().setLibrary()
3. Platform Testing ✅
- Shared code changes → Test iOS, Android, Web
- Check platform-specific gotchas in CLAUDE.md:
- Android: FlexWrap cards use 48% widths, font variants not fontWeight
- iOS: AsyncStorage debounced, NetInfo disabled
- Web: 3-column layout uses percentage widths, no Alert.alert
4. Deployment Process ✅
- ALWAYS update
firstPENDING_CHANGES.md - Use
./scripts/deploy.sh [qual|stage|beta|prod] - NEVER skip tests without explicit approval
- Run
before committingnpm run typecheck
5. Design Rules ✅
- NO GRAY TEXT - all text must be #000 (black)
- High contrast required for accessibility
- Use Typography component (handles font variants automatically)
Escalation Rules
Escalate to Higher Tier If:
- Quick → Iterative: Simple change but want validation
- Quick/Iterative → Standard: Multiple files affected, tests fail, edge cases emerge
- Standard → Full: 6+ files, security concerns, formal requirements needed
How to Escalate:
"Escalating to [TIER] workflow. [REASON: scope expanded, security implications, etc.]"
Then restart from Phase 1 of new tier.
Anti-Patterns (Never Do This)
❌ Use Quick workflow for new authentication system (use Full) ❌ Use Full workflow for fixing a typo (use Quick) ❌ Skip deployment phase to save time (tests are mandatory) ❌ Use
useAppStore.setState() directly (use store-specific methods)
❌ Use activity.name or activity.emoji (use text and icon)
❌ Gray text colors (use #000 only)
❌ Manual git commits (use deployment scripts)
Success Indicators by Tier
Quick Success:
- ✅ Change deployed in < 15 minutes
- ✅ Tests pass
- ✅ No rollbacks
Iterative Success:
- ✅ Change validated in < 30 minutes
- ✅ Peer review approved
- ✅ Tests pass
Standard Success:
- ✅ Feature complete in < 2 hours
- ✅ All edge cases covered
- ✅ Tests pass
- ✅ Peer review approved
Full Success:
- ✅ Epic complete with full documentation
- ✅ 100% acceptance criteria met
- ✅ Zero defects in production
- ✅ Full evidence trail
Resources
- Full decision matrix: See
resources/tier-selector.md - StackMap conventions: See project's
CLAUDE.md - Platform gotchas: See
directorydocs/
Example Orchestration
Example 1: Automatic Routing
User: "The sync system loses activity icons during conflict resolution" Your analysis: - Scope: Affects syncService.js, maybe conflict resolution logic (2-5 files) - Complexity: Needs research to understand conflict handling - Risk: Medium (data integrity issue) - Validation: Needs peer review → Decision: atlas-standard Your response: "I'll use the Atlas Standard workflow for this bug fix." [Invoke atlas-standard skill] [Execute 5-phase workflow]
Example 2: Explicit Routing
User: "Change primary button color to #007AFF. Use Atlas Quick." Your analysis: - Explicit tier specified: Quick - Verification: 1 file, trivial, zero risk ✅ → Decision: atlas-quick Your response: "Using Atlas Quick workflow for this trivial change." [Invoke atlas-quick skill] [Execute 2-phase workflow]
Example 3: Escalation
User: "Fix the modal padding issue" Your initial analysis: - Expected: 1 file, simple CSS change → Initial decision: atlas-quick During implementation: - Found: Affects 4 modal components + platform-specific styles - Complexity: Android uses different flex rules than iOS/Web - Risk: Could break modal layouts across platforms → Escalation: atlas-standard Your response: "Escalating to Standard workflow. Found 4 files affected with platform-specific considerations." [Invoke atlas-standard skill] [Restart from Phase 1: Research]
Integration with Agent Skills
Atlas includes specialized agent skills for specific phases:
- atlas-agent-developer: Implementation and planning (Sonnet)
- atlas-agent-peer-reviewer: Deep reviews, edge cases (Opus)
- atlas-agent-product-manager: Story creation, validation (Sonnet)
- atlas-agent-devops: Deployment, infrastructure (Sonnet)
- atlas-agent-security: Security audits (Sonnet)
These agents are invoked automatically during appropriate workflow phases.
Quick Start
As a User:
"[Describe your task]" → Orchestrator selects tier and executes workflow OR "[Task description]. Use Atlas [Quick|Iterative|Standard|Full] workflow." → Orchestrator uses specified tier
As the Orchestrator:
- Read task description
- Apply decision tree
- Invoke appropriate skill
- Execute workflow phases
- Apply StackMap-specific rules throughout
- Ensure quality gates pass before deployment
Remember: When in doubt, choose
atlas-standard. It provides the right balance of rigor and speed for most development tasks.