cd /news/ai-tools/compete-a-claude-code-plugin-for-int… · home topics ai-tools article
[ARTICLE · art-42566] src=github.com ↗ pub= topic=ai-tools verified=true sentiment=↑ positive

Compete – A Claude Code plugin for interactive competitor intelligence

Forthright Tech released Compete, a Claude Code plugin that automatically generates competitive intelligence reports from any software repository. The tool analyzes code, manifests, and README files to identify a product, then builds a multi-dimensional database covering competitors' product, technology, business, marketing, SEO, social media, customers, sales, hiring, funding, and brand positioning. It outputs a self-contained interactive HTML report with dashboards, comparison matrices, and prioritized recommendations, aiming to streamline competitive analysis for developers.

read6 min views1 publishedJun 28, 2026
Compete – A Claude Code plugin for interactive competitor intelligence
Image: source

A Claude Code Skill that turns your repository into a complete competitive intelligence report.

compete

analyzes the current repository to identify your product, then discovers competitors and builds a complete competitive intelligence database spanning product, technology, business, marketing, SEO, social media, customers, sales, hiring, funding, and brand positioning. It is an AI product research assistant — not just a comparison generator.

Everything is collected into normalized, confidence-annotated JSON datasets and rendered into a self-contained interactive HTML report (the InsightKit report) with dashboards, comparison matrices, capability radars, pricing ladders, a positioning scatter, SWOT accordions, and prioritized recommendations.

Repo-aware— reads your code, manifests, and README to identify the product before searching the web. No manual product brief required.Multi-dimensional intelligence— company profile, pricing, tech stack, social presence, marketing, and SEO for every competitor.** Confidence everywhere**— every field is wrapped with a confidence score, source, provenance, and an explicitunknown

fallback. Nothing is asserted without showing its work. Verify before acting.Normalized data contract— all stages write JSON validated against the schemas in, joined byskills/compete/schemas/

entity_ref

. Visualizations consume the data, never scrape directly.Self-contained report— onereport.html

file (~570 KB) opens standalone in any browser. The only external dependencies are the Chart.js and D3 CDN bundles.

A walkthrough of the seven tabbed views in a generated InsightKit report (this repository analyzed against its own competitive landscape):

compete

is packaged as a Claude Code plugin. The plugin bundles three things that work together:

  • the (compete

skillskills/compete/

) — the workflow, scripts, schemas, and report template; - the (/compete

slash commandcommands/compete.md

) — a one-shot entry point that drives the whole pipeline; - the plugin manifest ( .claude-plugin/plugin.json

).

/plugin marketplace add forthrighttech/compete
/plugin install compete

The first command registers this repository as a plugin marketplace (it ships a .claude-plugin/marketplace.json); the second installs the plugin. Restart or start a new session and confirm with

/plugin

compete

should be listed as enabled, the /compete

command available, and the compete

skill discoverable via /skills

.

git clone https://github.com/forthrighttech/compete.git \
  ~/.claude/plugins/compete

The plugin must live at a directory whose root contains .claude-plugin/plugin.json

(with skills/

and commands/

alongside it). Claude Code discovers it on the next session.

If you only want the skill without the slash command, clone just the skill subtree into your skills folder:

git clone https://github.com/forthrighttech/compete.git /tmp/compete && \
  cp -r /tmp/compete/skills/compete ~/.claude/skills/compete

cp -r /tmp/compete/skills/compete \
  /path/to/your-repo/.claude/skills/compete

The skill must live at .../.claude/skills/compete/

with SKILL.md

at its root. Confirm with /skills

(it should appear as compete

). You trigger it in natural language rather than with /compete

.

Claude Code with web access (WebSearch

/WebFetch

) for competitor discovery and intelligence collection.Python 3.9+ for the helper scripts in. The collection and rendering scripts use only the standard library — noskills/compete/scripts/

pip install

required.

Open Claude Code in the repository you want analyzed and run:

/compete

With no argument, Stage 1 auto-detects the product from the current repo and the command runs the full pipeline end to end, writing the report to ./insightkit-output/

.

You can also pass an optional seed — a competitor URL or name:

/compete https://www.crayon.co
/compete Klue

A seed is folded into Discovery as a known competitor candidate and is used to anchor the product's market/category, instead of relying solely on auto-detection. The seed is slugified into an entity_ref

(e.g. crayon

), classified, and appears in the roster alongside the auto-discovered competitors.

If you installed the skill without the command — or just prefer plain language — ask Claude directly. The skill triggers on phrases such as:

"Find my competitors" "Run a competitive analysis on this repo" "Who are my competitors and how do I compare?" "Build me a competitive landscape / positioning matrix / SWOT"

Claude runs the pipeline end to end. You can also drive any stage yourself:

python skills/compete/scripts/analyze_repo.py --repo . --validate

python skills/compete/scripts/discover_competitors.py plan  --product product.json
python skills/compete/scripts/discover_competitors.py build --product product.json \
  --candidates candidates.json --validate

python skills/compete/scripts/collect_intelligence.py plan  --competitors competitors.json
python skills/compete/scripts/collect_intelligence.py build --competitors competitors.json \
  --findings findings.json --validate

python skills/compete/scripts/build_report.py --input-dir . --output-dir ./insightkit-output

Open the result:

open ./insightkit-output/report.html

A fully rendered example ships in insightkit-output/ so you can see the report without running the pipeline:

File What it is
report.html

report.json

screenshots/overview.png

The sample analyzes this very repository (compete

) against 17 discovered competitors, classifying each by type and competitive threat. The report has seven tabbed views:

Overview— stat cards, competitor-classification and threat-distribution doughnuts, and the executive summary.** Comparison**— sortable matrix with confidence-bar cells per dimension.** Radar**— six transparent 0–100 capability axes, self vs. competitors, toggleable series.** Pricing**— entry-price bar chart plus a plan-ladder table.** Positioning**— D3 scatter of price × scale, bubble size = similarity, color = threat.** SWOT**— expandable strengths/weaknesses/opportunities/threats per competitor.** Opportunities**— market gaps and prioritized recommendations.

Every judgment in the report carries a method

note explaining the heuristic behind it.

repo ──▶ analyze_repo.py ──▶ product.json
                                 │
                                 ▼
        discover_competitors.py ──▶ competitors.json
                                 │
                                 ▼
        collect_intelligence.py ──▶ companies/pricing/techstack/
                                 │   social/marketing/seo.json
                                 ▼
            build_report.py ──▶ report.json ──▶ report.html

The normalized JSON datasets are the contract between stages. Each is validated against a schema in skills/compete/schemas/; the rules (confidence-wrapped fields,

unknown

fallback, entity_ref

joins) are documented in .

skills/compete/references/data-schema.md

Path Purpose
.claude-plugin/plugin.json

.claude-plugin/marketplace.json

/plugin install

.commands/compete.md

/compete

slash command (optional seed argument).skills/compete/SKILL.md

PRD.md

skills/compete/references/

skills/compete/scripts/

skills/compete/templates/

report.html

template.skills/compete/schemas/

insightkit-output/

compete

v1 ships the full one-shot pipeline: product intelligence → discovery → multi-dimensional collection → knowledge graph → interactive report. The following are deliberately deferred to v2.

v1 collects a solid breadth-first profile across all dimensions. v2 goes deep:

Deep SEO— keyword universe and rankings, backlink graph and authority, content gap analysis, SERP-feature ownership, and traffic-trend estimates (beyond v1's meta/structure snapshot).Deep social— engagement-rate and follower-growth time series, share-of- voice, sentiment, and channel-mix breakdowns per competitor.Hiring intelligence— open-roles tracking, team-growth and org-shape signals, and the strategic bets implied by what each competitor is hiring for.Sales intelligence— go-to-market motion (PLG vs. sales-led), funnel and packaging signals, win/loss themes, and target-segment inference.

v1 produces a point-in-time report. v2 adds a monitoring mode that re-runs the pipeline on a schedule and surfaces diffs — turning InsightKit from a one-time report generator into a continuous competitive-intelligence platform. Daily or weekly it would track:

  • pricing changes
  • new features
  • new blog posts
  • hiring trends
  • GitHub releases
  • social-media activity
  • funding news
  • SEO changes

See PRD.md for the complete v1 research scope and the v2 vision.

Contributions are welcome — see CONTRIBUTING.md for the data-contract rules, schema-validation workflow, and how to regenerate the sample report.

compete

is an open-source project built by forthrighttech, developed with the help of a few products I build:

— private AI that runs local LLMs on your machine.tokkong— an AI keyboard with on-device, local transcription.whiskey— a macOS menu bar manager, with support for macOS 27.lounge

Find me on X: @lbjhkg.

MIT © 2026 forthrighttech

── more in #ai-tools 4 stories · sorted by recency
── more on @forthright tech 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/compete-a-claude-cod…] indexed:0 read:6min 2026-06-28 ·