Market Recon — Community Research via MCP

This skill is for the main thread / orchestrator only. Subagents cannot call MCP tools directly. You fetch the data, write it to disk, and pass the file path to the subagent.

When to Use

DAG contains tasks with type=research AND signals like: community, sentiment, pain points, adoption, market, trends, opinions, practitioners, experience.

Query Strategy

Pick 2-3 subreddits matching the topic:
- Dev tooling / AI coding: ClaudeAI, ChatGPTCoding, cursor
- General dev: programming, webdev, ExperiencedDevs
- Infra / backend: devops, sysadmin, kubernetes
- Frontend: reactjs, nextjs, webdev
- Specific tech: use the technology's own subreddit
Search with short, specific queries (2-4 words). Run 2-3 queries per subreddit to cover angles: the tool name, the pain point, the alternative.
Skim results for signal: recurring complaints, praise patterns, specific user stories. Ignore hype posts with no substance.

Output Format

Write results to .godag/context/{task_id}-market.md:

# Market Recon: {topic}
## Sources
- r/{sub1}: {N} relevant posts searched
- r/{sub2}: {N} relevant posts searched

## Key Pain Points
1. {pain point} — {1-2 sentence evidence with post reference}
2. ...

## Positive Signals
1. {signal} — {evidence}

## Sentiment: {positive|negative|mixed|shifting}

## Gaps / Unmet Needs
- {gap}

Keep the file under 2000 tokens. The subagent needs a concise briefing, not a data dump.

Passing to Subagent

In the spawn prompt, add:

Read .godag/context/{task_id}-market.md for community research data.
Analyze and synthesize — don't just summarize what's in the file.