Startup Competitor Radar

Track what competitors change, why it matters, and what to do next.

Startup Competitor Radar is an open-source Codex skill for evidence-backed competitive intelligence. Give it your startup, product, URL, or competitor list. It researches the public web, builds a verified baseline, compares later snapshots, and produces an interactive Competitor Change Map instead of a generic competitor list.

It searches official websites, pricing and product pages, changelogs, documentation, GitHub, package registries, public X posts, search engines, news, public communities, review platforms, job listings, and available web archives.

What it does

• Resolves the correct company, domain, X handle, and GitHub organization before collecting evidence.
• Separates direct competitors, adjacent products, incumbents, and substitutes.
• Records source coverage, including blocked or unavailable sources.
• Saves immutable snapshots with stable signal keys.
• Detects pricing, packaging, positioning, product, integration, audience, hiring, and go-to-market movement.
• Verifies additions, removals, and modifications against comparable old and new evidence.
• Separates confirmed changes, corroborated signals, unconfirmed announcements, community reaction, and noise.
• Scores impact separately from confidence.
• Converts movement into four response lanes: Act, Investigate, Watch, Ignore.
• Generates a polished standalone HTML report and a traceable CSV evidence ledger.

Install

Install it globally for Codex with one command:

npx --yes startup-competitor-radar@latest

Or install directly from GitHub:

npx --yes skills add Kappaemme-git/startup-competitor-radar -g -y --skill track-startup-competitors --agent codex

Restart Codex after installation so the skill list refreshes.

Use it

Invoke the skill explicitly:

Use $track-startup-competitors to build a baseline for https://my-startup.example.
Track these competitors: Acme, Nova, and Flux.
Focus on pricing, positioning, product releases, and agency customers.

Or ask naturally:

Find what changed across my startup competitors in the last 30 days. Search their websites, X, GitHub, Google, news, communities, reviews, and public archives. Verify every important claim and tell me what to act on, investigate, watch, or ignore.

The skill always writes its research, labels, exports, and report in English.

Research modes

| Mode | Use it for | | --- | --- | | baseline | Create the first verified snapshot without pretending historical change is known. | | pulse | Find material public activity within a stated 7- or 30-day window. | | compare | Compare two snapshots and verify additions, removals, and modifications. | | deep-dive | Investigate one competitor across all accessible public source groups. | | market-shift | Find a supported movement pattern across several competitors. | | watchlist | Refresh an established competitor set and generate the full change map. |

Outputs

For a normal watchlist run, Codex creates:

competitor-radar/
└── snapshots/
    ├── 2026-07-01-watchlist.json
    └── 2026-08-01-watchlist.json

outputs/
├── startup-competitor-analysis.json
├── startup-competitor-change-map.html
├── startup-competitor-evidence.csv
└── startup-response-brief.md

The HTML report contains:

1. Executive market pulse
2. Source coverage and gaps
3. Filterable material changes
4. Movement timeline and cross-competitor themes
5. Act / Investigate / Watch / Ignore board
6. Competitor profiles
7. Uncertainty and counter-evidence
8. Filterable evidence ledger
9. Snapshot methodology

Evidence rules

The skill is deliberately conservative:

• Search-result snippets are discovery tools, never final proof.
• A feature announcement is not proof of availability, adoption, quality, or revenue.
• Missing current evidence is not proof that something was removed.
• X likes, GitHub stars, follower counts, and job listings are signals—not business results.
• A modified or removed signal requires comparable old and new evidence.
• Blocked, login-gated, or unavailable sources stay visible as coverage gaps.
• Every strategic interpretation is labeled as an inference and paired with uncertainty or counter-evidence.

It never bypasses authentication, private accounts, paywalls, CAPTCHAs, robots restrictions, or other access controls.

Try the fictional demo

The included OrbitDesk demo uses fictional companies and reserved .example URLs. It exists to test the pipeline safely; it is not real market research.

npm run demo

That command compares two snapshots, detects eight movements, generates the standalone report, and exports the evidence ledger.

Open:

outputs/orbitdesk-competitor-change-map.html

More demo prompts and the expected narrative are in DEMO.md.

Test

npm test

The tests cover deterministic snapshot comparison, report generation, CSV export, broken evidence references, misuse of search snippets, and installer configuration.

Project structure

track-startup-competitors/
├── SKILL.md
├── agents/openai.yaml
├── references/
│   ├── evidence-and-confidence.md
│   ├── identity-and-deduplication.md
│   ├── impact-scoring.md
│   ├── report-schema.md
│   ├── research-safety.md
│   ├── snapshot-schema.md
│   └── source-strategy.md
└── scripts/
    ├── compare_snapshots.mjs
    ├── export_evidence.mjs
    └── generate_report.mjs

Privacy

The workflow is public-source, read-only research. Snapshots must not store credentials, cookies, private messages, private profile content, or unrelated personal information. Review generated artifacts before sharing them outside your workspace.

License

MIT. See LICENSE.

100% open source.