Subtopics 6.1 through 6.4 covered the primitives: what a skill is, how to write one, what cron does, and how to chain them. Subtopic 6.5 is the worked example that ties them all together: a research sweep that runs every 6 hours, fans out to parallel sub-agents, applies a structured-output contract, and produces a gap-analysis brief that no single agent could produce alone.
The pattern is the channel's "Subagent Blueprint" video. The 6-hour cadence (0 */6 * * *) is the right interval for news and trend monitoring — frequent enough to catch emerging stories, sparse enough that each run has 15-30 new items to analyse (14-subagent-blueprint.md source). The skill is the recipe; the cron is the trigger; the sub-agents are the workers; the structured output is the contract; the gap analysis is the unique value.
This article is the longest in the course by design. The blueprint is the channel's most-viewed "production-ready" pattern, and treating it as a worked example means walking through every layer: the orchestrator, the sub-agents, the structured output, the synthesis, the gap analysis, the cron, the skill library, the failure modes, and the optimisation. The 5-hour refresh window is the budget; the 6-hour cadence is the heartbeat; the sub-agents are the muscle.
What you'll learn
- The orchestrator pattern: a main agent plans the research operation, spawns 4 parallel sub-agents (Twitter, news, Reddit, HN), collects structured outputs, synthesises findings, identifies gaps. The blueprint (14-subagent-blueprint.md source).
- The structured output contract: every sub-agent must return the same five sections — Key Findings, Notable Opinions, Source Links, Patterns Observed, Gaps Identified. The contract is what makes synthesis possible.
- The 6-hour cadence is optimal for news/trend monitoring. Hourly runs find <5 new items; 6-hour runs find 15-30; daily runs find 50+ but lose the "emerging trend" signal (14-subagent-blueprint.md source).
- Gap analysis is the unique value. Most research summarises what everyone is saying; the blueprint's value is finding what nobody is saying yet. Five gap types: underreported developments, missing perspectives, unanswered questions, emerging trends, contradictions.
- The skill library for the blueprint:
parallel_research(the orchestrator),twitter_intelligence,web_news_scraping,reddit_analysis,hn_analysis,identify_trends,gap_analysis,create_briefing_presentation,send_to_discord. Each is a leaf skill; the orchestrator chains them. - Sub-agents are role-names on a single model, not separate workers. The host's audit rule: name every sub-agent in the run log so you can tell which persona fired on which turn (g3adOMPsiiI, transcript, cross-listed from Course 4 §4.5).
- The 6-hour sweep costs ~30 prompts per run. A 24-hour day has 4 runs = 120 prompts. The 5-hour refresh window on a Minimax coding plan (100 prompts) can fit 3 runs; the Max tier (15,000) can fit unlimited. The 6-hour cadence is designed for the 5-hour window with one run to spare.
The orchestrator pattern
The blueprint's architecture (14-subagent-blueprint.md source):
┌─────────────────────────────────────┐
│ Main Agent (Orchestrator) │
│ - Plans research strategy │
│ - Spawns parallel sub-agents │
│ - Synthesises findings │
│ - Identifies gaps and patterns │
└──────────────┬──────────────────────┘
│
┌───────┴───────┬───────────┬──────────┐
│ │ │ │
┌──────▼──────┐ ┌─────▼──────┐ ┌──▼─────┐ ┌──▼─────┐
│ Sub-Agent 1 │ │ Sub-Agent 2│ │ Sub-A 3│ │ Sub-A 4│
│ Twitter/X │ │ News Sites │ │ Reddit │ │ HN │
│ Search │ │ Scraping │ │ Trends │ │ Posts │
└─────────────┘ └─────────────┘ └────────┘ └────────┘
│ │ │ │
└───────┬───────┴───────────┴──────────┘
│
┌───────▼────────────────────────────┐
│ Structured Output Collection │
│ - Key findings │
│ - Notable opinions │
│ - Source links │
│ - Patterns │
│ - Gaps │
└────────────────────────────────────┘
The orchestrator is the main agent. It plans the research operation, spawns the 4 sub-agents in parallel, collects the structured outputs, and runs the synthesis + gap analysis. The sub-agents are workers; they don't talk to each other; they report back to the orchestrator.
The orchestrator's prompt (the "parallel_research" skill body, simplified) (14-subagent-blueprint.md source):
You are the orchestrator and planner. Your job:
1. Plan the research operation
2. Spawn parallel sub-agents (Twitter, news, Reddit, HN)
3. Collect structured outputs from each sub-agent
4. Synthesise findings across sources
5. Identify gaps and opportunities
6. Save the final report to memory/research/YYYY-MM-DD-HH.md
7. Send summary to Discord #research channel
The 4 sub-agents run in parallel because they hit different sources with no inter-dependencies. The orchestrator waits for all 4 to complete (or timeout), collects the outputs, runs the synthesis, runs the gap analysis, saves the report.
The structured output contract
The contract is the load-bearing piece. Every sub-agent returns the same five sections (14-subagent-blueprint.md source):