competitive-intelligence-radar
CLAUDE.md4 starsChanged 5 months agoMITView on GitHub
Guidance for Claude Code when working in the Competitive Intelligence Radar plugin repository. A Claude Code plugin that runs a weekly competitor radar for business owners and convenes a strategic war room on demand. It showcases the BrightData MCP for live web data and deliberately uses BOTH Claude Code parallelism primitives, each where it honestly fits. - /radar-setup onboards: asks about and live-tests the user's BrightData and Notion connection, shows the proposed Notion schema for approval, checks agent-team prerequisites. -…
In this repository8 other files it gives its agents.
- radar-diff-changesSkill
- radar-merge-briefingSkill
- radar-onboardSkill
- radar-plan-targetsSkill
- radar-provision-notionSkill
- radar-publish-councilSkill
- radar-publish-notionSkill
- radar-render-dashboardSkill
The file
We split this CLAUDE.md into the rules it gives an agent
5 rules. About 0 are common, 2 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 22
If the user has just installed or cloned this, point them to /radar-setup.
- Distinctive: line 28
Do not turn the gather into an agent team.
- Distinctive: line 30
Do not turn the debate into plain subagents.
- Very specific: line 56
Always refer to BrightData and Notion tools BY CAPABILITY and build the full name from the prefix that radar-onboard detected and stored in …
- Very specific: line 86
Keep all user-specific data in radar.config.json so the shared plugin stays generic.
≈ “Common” is an estimate: at least 1 in 5 of the 28,029 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
CLAUDE.md
Guidance for Claude Code when working in the Competitive Intelligence Radar plugin repository.
What this project is
A Claude Code plugin that runs a weekly competitor radar for business owners and convenes a strategic war room on demand. It showcases the BrightData MCP for live web data and deliberately uses BOTH Claude Code parallelism primitives, each where it honestly fits.
/radar-setuponboards: asks about and live-tests the user's BrightData and Notion connection, shows the proposed Notion schema for approval, checks agent-team prerequisites./radaris the weekly GATHER: it fans out onecompetitor-scoutSUBAGENT per competitor in parallel; each gathers a cited dossier and reports back; the orchestrator diffs vs last run, merges one briefing, and publishes to three surfaces./radar-councilis the on-demand ANALYZE: it spins up a real experimental AGENT TEAM of four analyst teammates who message and challenge each other, then publishes a Strategic War Room addendum.
First thing to do in a fresh checkout
If the user has just installed or cloned this, point them to /radar-setup. Nothing else runs until
radar.config.json exists with verified MCP prefixes.
Architecture you must respect
- GATHER uses SUBAGENTS. Scouts are independent and only report back; they never message each other. Do not turn the gather into an agent team.
- DEBATE uses an AGENT TEAM. Analysts must see and challenge each other, which only a team allows. Do not turn the debate into plain subagents.
- The full rationale is in
rules/subagents-vs-agent-teams.md. Read it before changing either flow.
Hard rules (these are not optional)
These live in rules/. Every generated artifact must obey them:
no-fabrication.md: every factual claim carries a realsource_urlfetched this run plusfetched_at. No value without a source. Model knowledge is never a source.freshness.md: every field has a recency window and afetched_at; stale data is labeled, not presented as current.cost-control.md: prefer*_batchBrightData tools; cap competitors at 8; lean scout depth by default; the council is token heavy.automation-levels.md: AUTONOMOUS by default; CHECKPOINT for Notion database creation, for the MCP connection check at onboarding, for/radar-council, and for any GitHub push.no-em-dashes.md: never use the em dash (U+2014) or en dash (U+2013) anywhere, including in this file. Use commas, colons, or hyphens.
MCP tool resolution (important)
The same server resolves under different prefixes depending on connection method:
- Plugin path:
mcp__plugin_competitive-intelligence-radar_<server>__<tool> - Desktop app connectors:
mcp__claude_ai_<Server>__<tool> - Some sessions: a UUID prefix
mcp__<uuid>__<tool>
Always refer to BrightData and Notion tools BY CAPABILITY and build the full name from the prefix that
radar-onboard detected and stored in radar.config.json (mcp.brightdata_prefix,
mcp.notion_prefix). See references/brightdata-tool-map.md.
The agent-team caveat: when an agents/*.md definition runs as a TEAMMATE, its skills and
mcpServers frontmatter are NOT applied. Teammates load MCP and skills from the USER's settings. This
is why onboarding verifies the user's own connection. See references/agent-teams-setup.md.
Layout
.claude-plugin/ plugin.json + marketplace.json (repo is its own marketplace, source "./")
.mcp.json brightdata + notion, standard "mcpServers" wrapper (auto-loaded by the plugin)
commands/ /radar-setup, /radar, /radar-council (orchestrate only)
agents/ competitor-scout (subagent, sonnet) + 4 analyst teammate roles (sonnet)
skills/ radar-onboard, -provision-notion, -plan-targets, -diff-changes,
-merge-briefing, -publish-notion, -render-dashboard, -publish-council
rules/ the 6 hard rules above
references/ notion-schema, brightdata-tool-map, agent-teams-setup, config-schema
assets/ briefing-template.md, dashboard-template.html (self-contained, no deps)
outputs/ briefings/, dossiers/, dashboard/ (generated, gitignored)
radar.config.json the user's business context and run state (gitignored, never committed)Conventions when editing
- Reference bundled files from commands and skills with
${CLAUDE_PLUGIN_ROOT}. - Component directories live at the plugin ROOT, not under
.claude/. - The dashboard must stay self-contained: inline CSS and JS, data injected as inline JSON at the
{{RADAR_DATA_JSON}}placeholder, no external requests, no build step. - Keep all user-specific data in
radar.config.jsonso the shared plugin stays generic. - Skills are the workers; commands only sequence skills and (for
/radar) fan out subagents.
Verifying a change
node -e "JSON.parse(...)"the three JSON config files.- Grep the repo for U+2014 and U+2013; there should be zero hits.
- Confirm every dossier field is shaped
{ value, source_url, fetched_at, status }. - For flow changes, dry-run the relevant command and confirm all three surfaces still publish.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

