competitive-landscape

Competitive landscape

Map every direct and indirect competitor for the deal. Read lib/research-protocol.md and lib/playwright-auth.md before starting. Cap at ~5 fetches per competitor and ~30 competitors.

Inputs

Required (prompt if missing):

• Deal slug
• Company name
• One-line product description
• Target customer (ICP) if known

Pre-flight

Same idempotency check as other skills (read lib/deal.md). Artifact path: deals/<slug>/competitive-landscape.md. If it exists and --force was not passed, print "Artifact already exists at deals//competitive-landscape.md. Pass --force to overwrite." and stop.

Sources

1. Product Hunt — search the product category. Pull launch posts, traction signals, comments.
2. Crunchbase — Playwright with VC session. Funding history, headcount, last raise.
3. GitHub — search the category for open-source competitors. Star counts, commit cadence (active vs abandoned).
4. Public job boards — search incumbents for roles in this product area. Hiring signals indicate seriousness.
5. Google Patents — search the core technique or product noun. Filed-and-granted vs filed-and-abandoned matters.
6. News and tech press — search the category for the last 18 months.

Parallelization (Layer 2 — per competitor)

Competitor research splits cleanly per competitor. Run this skill in two stages:

Stage A — discovery (main session, ~10 fetches). Identify the candidate competitor list via batched parallel WebFetch (Layer 1) across Product Hunt, GitHub category search, and news/tech-press for the category. Output: a list of up to ~30 competitors with names + one-line positioning each.

Stage B — per-competitor deep dive (fan out). For the discovered list, dispatch one worker subagent per competitor, all concurrently in a single turn. See lib/research-protocol.md § Parallelization for the cross-platform mapping (Claude Code: Agent with subagent_type="general-purpose"; Codex: spawn_agent with agent_type="worker" and multi_agent = true in config).

Each subagent prompt MUST include:

• Competitor name, one-line positioning, and the deal's product description (so the subagent can judge direct vs indirect).
• A per-competitor budget of ~4 public-web fetches (reserve ~1 for main-session Crunchbase work).
• The citation and source-honesty rules from lib/research-protocol.md (paste, don't reference).
• Sources to consult: items 1, 3, 4, 5, 6 above. Skip item 2 (Crunchbase) — Playwright with VC session is main-session only.
• Required return shape: a single row each for the Direct-competitors table and (if applicable) the Indirect-competitors table from the artifact, plus any moat-relevant evidence quotes with sources, plus an Activity verdict (Active / Abandoned / Unknown) with last-commit or last-news date.

For incumbents specifically, the subagent additionally returns: shipping evidence link, hiring-signal count + link, public-statement quotes with sources.

After all subagents return, in the main session:

1. For each competitor that warrants it, do Crunchbase Playwright lookups (item 2) to fill funding/headcount/last-raise.
2. Merge subagent summaries + Crunchbase data into the artifact tables and the moat analysis.

If the discovered competitor list has fewer than 3 entries, skip fan-out and use Layer 1 inline.

Artifact template

# Competitive landscape: <Company>

**Deal:** <slug>
**Generated:** <ISO timestamp>

## Direct competitors

| Competitor | Positioning | Traction | Last funding | Moat type | Last activity | Source |
|------------|-------------|----------|--------------|-----------|---------------|--------|
| ... | ... | ... | ... | ... | ... | [link] |

## Indirect competitors

| Competitor | Why indirect | Risk it becomes direct | Source |
|------------|--------------|------------------------|--------|
| ... | ... | ... | [link] |

## Incumbent threat assessment

For each incumbent who could plausibly crush a startup in this space:

### <Incumbent name>

- **Currently shipping in this area?** Yes / No / Building. Evidence: [link]
- **Hiring signal?** [count of job postings tagged with the relevant keywords, with link]
- **Public statements?** [quotes from earnings calls, blog posts, conference talks, with sources]
- **Verdict:** Sleeping / Watching / Building / Already shipping.

## Moat analysis

For each candidate moat type, with evidence:

- **Network effects:** [analysis with evidence, or "Not applicable"]
- **Proprietary data:** [analysis with evidence, or "Not applicable"]
- **Switching costs:** [analysis with evidence, or "Not applicable"]
- **Brand:** [analysis with evidence, or "Not applicable"]

## Sources

- ...

## Open questions

- ...

After writing

Update deals/<slug>/manifest.json skills_completed["competitive-landscape"] to the current ISO-8601 UTC timestamp.