Radar Recommend — Personalized AI Recommendations

Radar Recommend — Personalized AI Recommendations

Match the scout catalogue against your personal context — goals, usage patterns, current projects, and installed tools — to surface what you should be paying attention to.

Arguments

• $ARGUMENTS — Optional:
  • --days N — How far back to look at session history (default: 14)
  • --focus <category> — Filter to a specific category (claude-code, mcp, api, agent-sdk, prompting, tooling, workflow, general-ai)

Parse from $ARGUMENTS if provided.

Workflow

Step 1: Load the Catalogue

Brain mode: Call get_lists and find all lists with names starting with [Radar] (excluding [Radar] Inbox). For each, call get_list and collect all items with status "open".

Local mode: Read ~/.claude/radar-catalogue.json and collect all items with status "open" across all lists (excluding Inbox).

If the catalogue is empty, tell the user: "No catalogue entries found. Run /radar-scan first to build your discovery catalogue."

If the catalogue file exists but cannot be parsed (corrupt JSON), print: "Catalogue file at [path] is corrupt. Delete it and re-run /radar-scan to rebuild."

If --focus was specified, filter items to only those with matching properties.category.

Step 2: Load Personal Context

Pull from multiple sources. Each is optional — work with whatever is available.

Brain goals: Call get_lists with pinned: true to get the user's stated goals and priorities. Extract goal titles and descriptions.

Brain thoughts: Call browse_recent with a generous limit (e.g., 50) to get recent thoughts. Note: browse_recent does not support date filtering — it returns the N most recent thoughts regardless of date. Filter results client-side by checking each thought's creation date, keeping only those from the last 14 days. Note recurring topics and themes.

Session history: Run: node "${CLAUDE_PLUGIN_ROOT}/bin/workflow-analyzer/dist/cli.js" parse --since ${DAYS} --output /tmp/discover-sessions.json

If the bundled binary is not available, fall back to npx @flippyhead/workflow-analyzer@latest parse --since ${DAYS} --output /tmp/discover-sessions.json.

Read the output file. If session history exceeds 50 sessions, summarize the top patterns:

• Most-used tools (top 10)
• Most-active projects (by session count)
• Recurring topics in user prompts
• Tool failure patterns

Current environment:

• Read ~/.claude/settings.json for installed permissions and allowed tools
• Look for .mcp.json files in the home directory and current project for installed MCP servers
• Check ~/.claude/plugins/ for installed plugins

Step 3: Match and Rank

For each open catalogue item, evaluate against the loaded context. Score on four dimensions:

Goal alignment (0-3):

• 3: Directly addresses a pinned goal
• 2: Related to a goal's domain
• 1: Tangentially useful
• 0: No connection

Usage gap (0-3):

• 3: User is doing something manually that this automates (evidence in session data)
• 2: User is using a tool that has a better/newer alternative
• 1: User works in the relevant domain but hasn't needed this yet
• 0: No gap identified

Recency (0-2):

• 2: Released in the last 7 days
• 1: Released in the last 30 days, or newly relevant due to a recently started project
• 0: Older

Effort/impact (0-2):

• 2: Low effort, high impact (e.g., a config change or install command)
• 1: Medium effort or medium impact
• 0: High effort or low impact

Total score: 0-10. Skip items scoring below 3 — they don't connect to the user's context meaningfully.

Step 4: Present Recommendations

Sort by total score descending. Group into tiers:

Act Now (score 7-10): Items with high relevance and low effort. Lead with what the user is doing that this improves. Format:

[Title] (score: N/10) You're [specific observation from session data or goals]. [This tool/feature] [specific benefit]. Next step: [concrete action — install command, link to try, config change]

Worth Exploring (score 5-6): Items with high relevance but higher effort. Format:

[Title] (score: N/10) Given your goal of [goal], this [what it does]. Worth a deeper look when [suggested timing]. Link: [url]

On Your Radar (score 3-4): Items with moderate relevance. Brief format:

[Title] — [one sentence on what it is and why it might matter] ([url])

Limit output to:

• Act Now: up to 5 items
• Worth Exploring: up to 5 items
• On Your Radar: up to 5 items

If no items score above 3, report: "Nothing in the current catalogue connects strongly to your goals and usage patterns. The catalogue may need more entries — try running /radar-scan or adding items to [Radar] Inbox."

Step 5: Publish as Insights

Brain mode: Publish the recommendations as insights using create_report so they appear in the AI Brain insights UI. This is the primary output of discover — not just terminal text.

Call create_report with:

• startDate / endDate: the period covered by session history
• sessionsAnalyzed, totalPrompts, totalToolCalls: from parsed session data (use 0 if session history unavailable)
• projectsActive: from session data (use empty array if unavailable)
• modelUsage: from session data (use empty object if unavailable)
• insights: array of recommendations, each with:
  • category: use "feature-discovery" for Act Now and Worth Exploring items, "ecosystem" for broader tools/techniques
  • observation: what the data shows — cite the specific goal, usage pattern, or environment detail that triggered the match
  • recommendation: the concrete action to take
  • evidence: include the score breakdown (goal alignment, usage gap, recency, effort/impact) and total score
  • links: array of {label, url} for the catalogue item URL and any related resources

Only publish items scoring 5+ (Act Now and Worth Exploring tiers). On Your Radar items are too low-signal for the insights UI — just mention them in the terminal summary.

Terminal-only mode: If brain MCP tools are unavailable, skip publishing. The terminal output from Step 4 is the primary output. Do not warn about brain being unavailable.

Step 6: Update Catalogue State

For each item that was recommended, update its properties to include:

• lastRecommended: today's ISO date
• matchedGoals: array of goal titles it matched against
• matchedPatterns: array of usage patterns that triggered the match

Brain mode: The update_list_item MCP tool replaces the entire properties field — it does NOT merge. So you must: (1) read the item's existing properties from the get_list response, (2) merge the new keys (lastRecommended, matchedGoals, matchedPatterns) into the existing properties object client-side, (3) send the full merged object to update_list_item. Local mode: Update the JSON file (same merge-then-write approach).

Items with lastRecommended within the last 14 days should be deprioritized (reduce score by 2) on subsequent runs to avoid re-surfacing the same recommendations.

Step 7: Summary

Output a brief terminal summary:

• How many insights were published to the brain
• The top 2-3 "Act Now" recommendations (one line each)
• If brain is connected: "Recommendations also saved to your brain. Review at /insights in the AI Brain web UI."
• If terminal-only: no brain reference in the output