Agent Competitive Analysis

The output is a decision, not a catalog: where do we differentiate, and where do we just need parity? Everything else serves that.

Conventions

Follow _conventions.md (read-first order, docs/ vs artifacts/, don't-clobber guard).

Before you start

Read PROJECT.md and artifacts/prd.md. Anchor on the core task and the user's job-to-be-done — compare on that, not on feature checklists.

Gathering information

If web search is available, use it to find current competitors and recent moves — agent/AI tooling shifts fast, so prefer recent sources and note the date. If web search is not available, work from what the user knows and clearly mark gaps as "needs verification" rather than guessing. Don't invent competitor facts.

What to compare on

For each competitor: how they solve the core job, their autonomy level (do they take actions or just suggest?), where they're strong, where they're weak/annoying, and pricing/positioning if known. For an agent product specifically, note trust and reliability posture — how they handle being wrong — since that's often the real battleground, not features.

Output

Write artifacts/competitive.md (apply the don't-clobber guard):

# Competitive brief — [product]
## The job we're competing on
## Landscape (2–6 competitors)
For each: approach · autonomy · strengths · weaknesses · trust posture · pricing
## Gaps in the market
Where is everyone weak? That's the opening.
## Differentiate vs. parity
- Differentiate on: (the 1–2 things we'll be clearly best at)
- Parity on: (table stakes we just can't be bad at)
- Ignore for now: (things that look important but aren't, for our wedge)
## Positioning one-liner
## Sources & confidence

End with the single sharpest strategic implication for the build — usually "differentiate on X, don't waste effort matching Y."