Initial commit

2025-11-30 08:53:07 +08:00
commit b5ab310674
19 changed files with 5687 additions and 0 deletions
--- a/skills/agent-delegation/examples/parallel-research.md
+++ b/skills/agent-delegation/examples/parallel-research.md
@@ -0,0 +1,337 @@
+# Example: Parallel Research Delegation
+
+This example shows how to decompose a research task into parallel specialist activities.
+
+## Scenario
+
+**User Request:** "Research competitive landscape for our B2B SaaS pricing strategy"
+
+## Task Decomposition
+
+```
+Original Task: Research competitive landscape for B2B SaaS pricing
+
+Activities Identified:
+1. Analyze Competitor A (Salesforce) pricing
+   - Expertise: Market research
+   - Output: Pricing analysis document
+   - Dependencies: None
+
+2. Analyze Competitor B (HubSpot) pricing
+   - Expertise: Market research
+   - Output: Pricing analysis document
+   - Dependencies: None
+
+3. Analyze Competitor C (Zendesk) pricing
+   - Expertise: Market research
+   - Output: Pricing analysis document
+   - Dependencies: None
+
+4. Synthesize competitive findings
+   - Expertise: Strategic analysis
+   - Output: Unified competitive strategy
+   - Dependencies: Activities 1, 2, 3
+
+Execution Strategy: Mixed
+- Parallel: Activities 1-3 (independent research)
+- Sequential: Activity 4 (requires all results)
+
+Reasoning: Each competitor analysis is independent. Synthesis must wait for all analyses to complete.
+```
+
+## Agent Prompts Generated
+
+### Agent 1: Salesforce Analysis
+
+```
+FOCUS: Research Salesforce's B2B SaaS pricing strategy
+    - Identify all pricing tiers (Essentials, Professional, Enterprise, Unlimited)
+    - Map features to each tier
+    - Note promotional strategies and discounts
+    - Calculate price-per-user for each tier
+    - Identify what features drive tier upgrades
+
+EXCLUDE:
+    - Don't analyze their technology stack or implementation
+    - Don't make pricing recommendations yet (synthesis phase)
+    - Don't compare to other competitors (separate analyses)
+    - Don't analyze Salesforce Platform (focus on CRM product)
+
+CONTEXT: We're researching competitive landscape for our B2B SaaS pricing strategy.
+    - Target market: Mid-market companies (100-500 employees)
+    - Looking for pricing patterns and positioning insights
+    - Focus on transparent, public pricing only
+
+OUTPUT: Structured analysis document including:
+    - Pricing Tiers Table (tier name, price, billing cycle)
+    - Feature Matrix by Tier (features mapped to each tier)
+    - Key Insights (3-5 bullet points about their strategy)
+    - Notable Patterns (what we can learn from their approach)
+    - Competitive Positioning (how they position each tier)
+
+SUCCESS: Complete analysis with actionable data
+    - All publicly available tiers documented
+    - Features accurately mapped to tiers
+    - Insights are specific and evidence-based
+    - Patterns identified are relevant to our market
+
+TERMINATION:
+    - Analysis complete with all sections filled
+    - Information not publicly available (document what's missing)
+    - Maximum 2 hours of research time reached
+```
+
+### Agent 2: HubSpot Analysis
+
+```
+FOCUS: Research HubSpot's B2B SaaS pricing strategy
+    - Identify all pricing tiers (Starter, Professional, Enterprise)
+    - Map features to each tier
+    - Note freemium offering and upgrade triggers
+    - Calculate price-per-contact for marketing hub
+    - Identify bundling strategy across products
+
+EXCLUDE:
+    - Don't analyze technology implementation
+    - Don't make pricing recommendations yet
+    - Don't compare to other competitors
+    - Don't analyze HubSpot Academy or services pricing
+
+CONTEXT: Competitive landscape research for B2B SaaS pricing.
+    - Target market: Mid-market companies (100-500 employees)
+    - Interest in freemium-to-paid conversion strategies
+    - Focus on Marketing Hub and Sales Hub pricing
+
+OUTPUT: Structured analysis document including:
+    - Pricing Tiers Table
+    - Feature Matrix by Tier
+    - Freemium Analysis (what's free, what triggers upgrade)
+    - Key Insights (3-5 bullet points)
+    - Bundling Strategy (how products bundle together)
+    - Notable Patterns
+
+SUCCESS: Complete analysis with freemium insights
+    - All tiers documented
+    - Freemium model understood
+    - Bundle pricing analyzed
+    - Insights are actionable
+
+TERMINATION:
+    - Analysis complete
+    - Information unavailable
+    - 2 hours research time reached
+```
+
+### Agent 3: Zendesk Analysis
+
+```
+FOCUS: Research Zendesk's B2B SaaS pricing strategy
+    - Identify all pricing tiers (Suite Team, Growth, Professional, Enterprise)
+    - Map features to each tier
+    - Note per-agent pricing model
+    - Analyze add-on pricing strategy
+    - Identify what drives customer to higher tiers
+
+EXCLUDE:
+    - Don't analyze implementation details
+    - Don't make recommendations yet
+    - Don't compare to competitors
+    - Don't analyze Zendesk Sell (focus on Support product)
+
+CONTEXT: Competitive research for B2B SaaS pricing strategy.
+    - Target market: Mid-market companies
+    - Interest in per-seat vs per-usage pricing models
+    - Focus on customer support software pricing
+
+OUTPUT: Structured analysis including:
+    - Pricing Tiers Table
+    - Feature Matrix by Tier
+    - Per-Agent Economics (cost per agent at scale)
+    - Add-On Pricing (what extras cost money)
+    - Key Insights (3-5 bullet points)
+    - Upgrade Drivers (what makes customers move up tiers)
+
+SUCCESS: Complete analysis with per-seat insights
+    - All tiers documented
+    - Per-agent model understood
+    - Add-on strategy mapped
+    - Insights are specific
+
+TERMINATION:
+    - Analysis complete
+    - Information unavailable
+    - 2 hours research reached
+```
+
+### Agent 4: Synthesis (Sequential, after 1-3)
+
+```
+FOCUS: Synthesize competitive pricing findings into unified strategy insights
+
+EXCLUDE:
+    - Don't create our pricing (separate decision)
+    - Don't add new competitor research
+    - Don't analyze product features (stick to pricing patterns)
+
+CONTEXT: We have three completed competitor analyses.
+    - Salesforce analysis: [Results from Agent 1]
+    - HubSpot analysis: [Results from Agent 2]
+    - Zendesk analysis: [Results from Agent 3]
+    - Goal: Extract patterns and strategic insights
+
+OUTPUT: Competitive Pricing Strategy Document including:
+    - Cross-Competitor Patterns (themes across all three)
+    - Pricing Model Comparison (per-user vs per-contact vs per-usage)
+    - Tier Strategy Analysis (how competitors structure tiers)
+    - Freemium Insights (who uses it, how it works)
+    - Feature Differentiation (what features separate tiers)
+    - Positioning Opportunities (gaps or opportunities for us)
+    - Strategic Recommendations (3-5 actionable insights)
+
+SUCCESS: Unified synthesis with strategic value
+    - Patterns identified across competitors
+    - Comparisons are fair and evidence-based
+    - Recommendations are actionable
+    - No contradiction between analyses
+
+TERMINATION:
+    - Synthesis complete
+    - Conflicting data needs clarification (escalate)
+    - All sections filled
+```
+
+## Execution Flow
+
+### Phase 1: Launch Parallel Research (Agents 1-3)
+
+```
+🚀 Launching 3 parallel research agents
+
+Agent 1: Salesforce Analysis → RUNNING
+Agent 2: HubSpot Analysis → RUNNING
+Agent 3: Zendesk Analysis → RUNNING
+
+File Coordination Check:
+✅ Agent 1 OUTPUT: research/salesforce-pricing.md
+✅ Agent 2 OUTPUT: research/hubspot-pricing.md
+✅ Agent 3 OUTPUT: zendesk-pricing.md
+✅ All paths unique - No collision risk
+
+Parallel Safety Check:
+✅ Independent tasks - No dependencies
+✅ Read-only research - No shared state
+✅ Separate validation - Each can be reviewed independently
+✅ Won't block - Different data sources
+
+Status: ✅ SAFE FOR PARALLEL EXECUTION
+```
+
+### Phase 2: Await Completion
+
+```
+⏳ Waiting for parallel agents...
+
+Agent 1: COMPLETE ✅ (45 minutes)
+  - Salesforce analysis delivered
+  - All sections complete
+  - 4 tiers documented
+
+Agent 2: COMPLETE ✅ (52 minutes)
+  - HubSpot analysis delivered
+  - Freemium model analyzed
+  - 3 tiers + free tier documented
+
+Agent 3: COMPLETE ✅ (38 minutes)
+  - Zendesk analysis delivered
+  - Per-agent model mapped
+  - 4 tiers + add-ons documented
+
+All parallel agents complete. Proceeding to synthesis.
+```
+
+### Phase 3: Launch Synthesis (Agent 4)
+
+```
+🚀 Launching synthesis agent
+
+Agent 4: Competitive Synthesis → RUNNING
+
+Context provided:
+- Salesforce results (from Agent 1)
+- HubSpot results (from Agent 2)
+- Zendesk results (from Agent 3)
+
+Agent 4: COMPLETE ✅ (25 minutes)
+  - Cross-competitor patterns identified
+  - 5 strategic recommendations generated
+  - Positioning opportunities highlighted
+```
+
+## Results
+
+### Total Time: 52 minutes (parallel) + 25 minutes (synthesis) = 77 minutes
+
+**Compare to sequential:** 45 + 52 + 38 + 25 = 160 minutes
+**Time saved:** 83 minutes (52% faster)
+
+### Deliverables
+
+```
+📁 research/
+├── salesforce-pricing.md (Agent 1)
+├── hubspot-pricing.md (Agent 2)
+├── zendesk-pricing.md (Agent 3)
+└── competitive-strategy.md (Agent 4 synthesis)
+```
+
+### Key Insights Generated
+
+From the synthesis agent:
+
+1. **Tiering Pattern:** All three use 3-4 tier structure with similar progression (basic → professional → enterprise)
+
+2. **Pricing Models:** Mixed approaches
+   - Salesforce: Per-user, all-inclusive features
+   - HubSpot: Per-contact, freemium base
+   - Zendesk: Per-agent, add-on marketplace
+
+3. **Feature Gating:** Core features in all tiers, advanced analytics/automation in top tiers
+
+4. **Freemium:** Only HubSpot uses freemium successfully (strong upgrade triggers identified)
+
+5. **Opportunity:** Gap in mid-market transparent pricing - competitors hide "contact sales" behind top tier
+
+## Lessons Learned
+
+### What Worked Well
+
+✅ **Parallel execution:** Saved 52% time
+✅ **Independent research:** No coordination overhead
+✅ **Synthesis phase:** Unified findings effectively
+✅ **Unique file paths:** No collisions
+✅ **Explicit FOCUS/EXCLUDE:** Agents stayed on task
+
+### Improvements for Next Time
+
+- Add time limits to prevent research rabbit holes
+- Specify exact format (all agents used slightly different table formats)
+- Request specific pricing data points (some agents missed cost-per-user calculations)
+- Consider adding validation agent before synthesis (check data accuracy)
+
+## Reusable Template
+
+This pattern works for any parallel research:
+
+```
+1. Decompose research into independent topics
+2. Create identical FOCUS/EXCLUDE templates
+3. Customize context and output paths only
+4. Launch all in parallel
+5. Synthesis agent consolidates findings
+```
+
+**Use when:**
+- Researching multiple competitors
+- Analyzing multiple technologies
+- Gathering multiple data sources
+- Interviewing multiple stakeholders