@garethdaine/agentops 0.9.3 → 0.10.0

package/README.md

@@ -6,7 +6,7 @@
 
 A plugin for [Claude Code CLI](https://docs.anthropic.com/en/docs/claude-code) that wraps every session in 7 security layers, structures work with the STAR methodology, auto-pilots workflows, delegates to 12 specialist agents, learns from failures via self-evolution, and orchestrates full project builds from vision to merged PR.
 
-37 slash commands | 44 hooks | 12 specialist agents | 49 templates | 32+ feature flags
+40 slash commands | 44 hooks | 12 specialist agents | 49 templates | 33+ feature flags | real-time 3D dashboard
 
 [![Tests](https://github.com/garethdaine/agentops/actions/workflows/tests.yml/badge.svg?branch=main)](https://github.com/garethdaine/agentops/actions/workflows/tests.yml)
 [![Lint](https://github.com/garethdaine/agentops/actions/workflows/lint.yml/badge.svg?branch=main)](https://github.com/garethdaine/agentops/actions/workflows/lint.yml)
@@ -44,7 +44,7 @@ On first session, the plugin auto-initializes `.agentops/` with default flags an
 | Dimension | AgentOps | GSD (~31K stars) | Superpowers (~50K stars) |
 |-----------|----------|-------------------|--------------------------|
 | **Core identity** | Enterprise guardrailing + full delivery lifecycle | Spec-driven meta-prompting to beat context rot | Skills-based methodology with TDD enforcement |
-| **Commands** | 37 | ~15 | ~10 (skill-based) |
+| **Commands** | 40 | ~15 | ~10 (skill-based) |
 | **Hooks** | 44 shell scripts across 7 lifecycle events | None (prompts only) | None (prompts only) |
 | **Agents** | 12 specialist agents | 4 parallel researchers + planner | Code reviewer agent |
 | **Security** | 7 layers (injection, exfiltration, supply-chain, Unicode, credential, path, env) | Minimal | None |
@@ -61,6 +61,7 @@ On first session, the plugin auto-initializes `.agentops/` with default flags an
 | **Observability** | Audit logs, OTLP telemetry, file provenance, cost budgets | None | None |
 | **Configurable autonomy** | guided / supervised / autonomous | N/A | N/A |
 | **Linear integration** | Task sync (create, status update, close) | None | None |
+| **Real-time dashboard** | 3D office scene (React Three Fiber) + 2D fallback, WebSocket relay | None | None |
 
 ### What AgentOps does that neither competitor does
 
@@ -68,6 +69,7 @@ On first session, the plugin auto-initializes `.agentops/` with default flags an
 2. **EvoSkill self-evolution.** The failure-collector → proposer → skill-builder → feedback-history pipeline auto-generates skills from failures. No manual intervention.
 3. **Full observability.** Structured audit logs, OTLP telemetry export, file provenance tracking, and cost budgets with session-level granularity.
 4. **8.5-phase build lifecycle.** From brainstorm to merged PR with state machine, resumability, human gates, parallel research, TDD enforcement, two-stage review, and Nyquist verification — all configurable via feature flags and autonomy levels.
+5. **Real-time 3D agent dashboard.** A Next.js 16 + React Three Fiber office scene that visualizes agent activity live via WebSocket relay from telemetry JSONL, with 2D fallback for non-GPU environments.
 
 ---
 
@@ -78,7 +80,7 @@ The plugin integrates through four extension points:
 | Extension Point | Location | Count | Purpose |
 |-----------------|----------|-------|---------|
 | **Hooks** | `hooks/hooks.json` | 44 | Intercept tool use at every lifecycle event |
-| **Commands** | `commands/*.md` | 37 | User-facing slash commands (`/agentops:*`) |
+| **Commands** | `commands/*.md` | 40 | User-facing slash commands (`/agentops:*`) |
 | **Agents** | `agents/*.md` | 12 | Specialist subagents for analysis and execution |
 | **Templates** | `templates/**/*.md` | 49 | Standards, architecture patterns, delivery docs |
 
@@ -134,7 +136,7 @@ Plus supply-chain defense: Unicode/Glassworm detection (`unicode-firewall.sh`),
 
 ---
 
-## Commands (37)
+## Commands (40)
 
 ### Core Commands
 
@@ -181,6 +183,9 @@ Plus supply-chain defense: Unicode/Glassworm detection (`unicode-firewall.sh`),
 | `/agentops:dev-setup` | Developer environment setup guide |
 | `/agentops:docker-dev` | Docker development environment configuration |
 | `/agentops:e2e` | End-to-end test planning and execution |
+| `/agentops:worklog` | Personal work log for tracking daily contributions |
+| `/agentops:idea` | Structured idea proposals with development and export |
+| `/agentops:search` | Unified search across work log, ideas, git history, and knowledge base |
 | `/agentops:herd` | Multi-agent coordination for complex tasks |
 
 ---
@@ -266,7 +271,7 @@ Standards are enforced via `templates/standards/standards-checklist.md` during P
 
 ---
 
-## Feature Flags (32+)
+## Feature Flags (33+)
 
 All flags default to sensible values and are toggleable via `/agentops:flags` or `.agentops/flags.json`.
 
@@ -282,6 +287,9 @@ All flags default to sensible values and are toggleable via `/agentops:flags` or
 ### Build lifecycle flags
 `build_tdd_enforced`, `build_parallel_research`, `build_xml_plans`, `build_linear_sync`, `build_fresh_context`, `build_wave_parallel`, `build_nyquist_enforce`, `build_persuasion`, `build_quick_mode`, `build_scaffold_auto`, `build_standards_inject`, `standards_enforcement_mode`
 
+### Dashboard flags
+`dashboard_enabled`
+
 ### Enterprise flags
 `enterprise_scaffold`, `ai_workflows`, `unified_review`, `architecture_guardrails`, `delivery_lifecycle`, `team_governance`, `client_comms`
 
@@ -314,6 +322,37 @@ All flags default to sensible values and are toggleable via `/agentops:flags` or
 
 ---
 
+## Agent Office Dashboard
+
+A real-time 3D visualization of agent activity, built with Next.js 16, React Three Fiber, and Shadcn UI.
+
+### Features
+
+| Feature | Description |
+|---------|-------------|
+| **3D office scene** | Workstations with lighting, agent avatars, and animations via React Three Fiber |
+| **Real-time state** | WebSocket relay server streams agent activity from JSONL telemetry files |
+| **2D fallback** | Activity table with WebGL detection for environments without GPU support |
+| **Session registry** | Auto-registers sessions from incoming telemetry with stale detection |
+| **Connection status** | Live indicator showing relay connection health |
+| **Lifecycle management** | Proper Three.js disposal and resource cleanup |
+
+### How it works
+
+```
+hooks/telemetry.sh → .agentops/telemetry.jsonl
+        ↓
+dashboard/server/file-watcher.ts (byte-offset tracking)
+        ↓
+dashboard/server/relay.ts (WebSocket with origin validation)
+        ↓
+dashboard/src/ (Zustand store → 3D scene / 2D table)
+```
+
+The dashboard launches automatically when `dashboard_enabled` is `true` in `.agentops/flags.json`. The `dashboard-launch.sh` hook starts the Next.js dev server and WebSocket relay on session start.
+
+---
+
 ## Configuration
 
 ### Environment Variables
@@ -376,7 +415,11 @@ agentops-plugin/
 │   ├── communication/              # Stakeholder comms templates
 │   ├── workflows/                  # Feature, refactor, spike, bug workflows
 │   └── scaffolds/                  # Error handling, logging, health checks
-├── tests/                          # BATS test suite
+├── dashboard/                      # Agent Office Dashboard (Next.js 16)
+│   ├── src/                        # React components, stores, types
+│   ├── server/                     # WebSocket relay + JSONL file watcher
+│   └── tests/                      # Dashboard unit tests (Vitest)
+├── tests/                          # 20 BATS test files (security, automation, dashboard)
 ├── docs/                           # Architecture docs
 ├── settings.json                   # Plugin permission defaults
 ├── LICENSE                         # MIT
@@ -387,18 +430,37 @@ agentops-plugin/
 
 ## Testing
 
-The plugin includes a BATS test suite covering security hooks:
+The plugin includes 20 BATS test files covering security hooks, automation hooks, and dashboard integration:
 
 ```bash
 # Run all tests
 bats tests/
 
-# Run specific test file
+# Security hooks
 bats tests/validate-command.bats
+bats tests/validate-path.bats
+bats tests/validate-env.bats
 bats tests/injection-scan.bats
 bats tests/exfiltration-check.bats
-bats tests/validate-path.bats
+bats tests/credential-redact.bats
+bats tests/content-trust.bats
 bats tests/feature-flags.bats
+
+# Automation hooks
+bats tests/auto-test.bats
+bats tests/auto-plan.bats
+bats tests/auto-verify.bats
+bats tests/auto-lesson.bats
+bats tests/auto-evolve.bats
+bats tests/auto-delegate.bats
+bats tests/failure-collector.bats
+bats tests/evolve-gate.bats
+
+# Dashboard
+bats tests/dashboard-flag.bats
+bats tests/dashboard-launch.bats
+bats tests/session-registry.bats
+bats tests/telemetry-pretool.bats
 ```
 
 ---

package/commands/enterprise/idea.md

@@ -0,0 +1,188 @@
+---
+name: idea
+description: Develop, document, and export professional idea proposals with structured thinking and shareable deliverables
+---
+
+You are an AI-powered idea development assistant. You help capture, develop, and share professional idea proposals with configurable depth.
+
+**Before starting, check the feature flag:**
+Run: `source hooks/feature-flags.sh && agentops_enterprise_enabled "work_journal"` — if disabled, inform the user: "Work journal commands are disabled. Enable with: /agentops:flags work_journal true" and stop.
+
+## CRITICAL RULE: Use AskUserQuestion Tool
+
+You MUST use the `AskUserQuestion` tool for EVERY question in this command. DO NOT print questions as plain text or numbered option lists. Call the AskUserQuestion tool which renders a proper selection UI. This is a BLOCKING REQUIREMENT.
+
+**Read the autonomy level** from `.agentops/flags.json` (key: `autonomy_level`). Default to `guided` if not set.
+
+---
+
+## Startup
+
+1. **Read conventions:** Read `templates/work-journal/conventions.md` for shared schemas, storage layout, index format, integration detection, and templates. Follow all protocols defined there.
+2. **Load config:** Read `.agentops/journal-config.json`. If it does not exist, run the First-Run Setup defined in conventions.md — use `AskUserQuestion` for each step (name, role, team, contacts, integrations). Write the resulting config file.
+3. **Detect integrations:** Run the Integration Detection protocol from conventions.md once. Cache results. Report: "Integrations: Cortex [status], Notion [status]" or "Running in local-only mode."
+4. **Check for existing idea:** If `$ARGUMENTS` contains a slug matching an existing idea in `.agentops/journal/ideas/`, go to the **Revisit Existing Idea** section instead.
+
+---
+
+## Capture Phase — New Idea
+
+Use `AskUserQuestion` for each of these in sequence:
+
+1. **Idea description:** Free text question — "Describe your idea. What problem does it solve and what do you propose?"
+2. **Category:** "What category best fits this idea?" — options: "Architecture improvement", "Process improvement", "New tool or automation", "Client delivery approach", "Cost reduction", "Quality improvement", "Team workflow", "Business opportunity", "Other"
+3. **Beneficiaries:** "Who benefits most from this idea?" — options: "Engineering team", "Client", "Business", "End users", "Multiple stakeholders"
+4. **Development depth:** "How deeply should we develop this idea?" — options:
+   - "Quick (5 min) — One-page proposal with core elements"
+   - "Light (15 min) — Analysis with alternatives, trade-offs, and risks"
+   - "Full (30+ min) — Research-backed proposal with implementation plan and metrics"
+
+Generate a kebab-case slug from the idea title. Create the directory: `.agentops/journal/ideas/{slug}/` using `mkdir -p`. Create `.agentops/journal/ideas/{slug}/scratchpad.md` for working notes.
+
+---
+
+## Quick Capture (5 min)
+
+Generate a one-page proposal document at `.agentops/journal/ideas/{slug}/proposal.md` using the Idea Proposal frontmatter and document structure from conventions.md. Include these sections:
+
+- **Problem Statement** — What problem does this solve?
+- **Proposed Solution** — What is being proposed?
+- **Expected Benefits** — What improves?
+- **Recommendation** — Effort estimate and specific next step
+
+Set status to `Draft`. Sign with author name and role from `journal-config.json` identity. Timestamp with current UTC date.
+
+After writing, proceed to **Post-Capture** section.
+
+---
+
+## Light Development (15 min)
+
+Complete everything in Quick Capture, then expand the proposal:
+
+1. **Structured analysis:** Use `/agentops:reason` to analyse the idea — evaluate feasibility, identify assumptions, and surface risks.
+2. **Knowledge lookup:** Cross-reference `/agentops:knowledge` to check for related patterns, prior decisions, or existing ADRs.
+3. **Expand sections:** Fill in Alternatives Considered, Trade-offs & Risks, and Implementation Sketch sections in the proposal.
+
+Update the proposal document with all expanded sections. Proceed to **Post-Capture**.
+
+---
+
+## Full Development (30+ min)
+
+Complete everything in Light Development, then deepen with research:
+
+1. **Web research:** Search the web for prior art, industry practices, benchmarks, and supporting data relevant to this idea. Save all research notes to `.agentops/journal/ideas/{slug}/research.md` with source URLs and key findings.
+2. **Detailed implementation plan:** Add phased implementation approach with effort estimates, dependencies, and milestones to the Implementation Sketch section.
+3. **Cost-benefit analysis:** Fill in the Cost-Benefit Assessment section with time/effort investment vs expected return, including quantitative estimates where possible.
+4. **Success metrics:** Define measurable success criteria in the Success Metrics section.
+
+Update the proposal document with all sections complete. Proceed to **Post-Capture**.
+
+---
+
+## Post-Capture
+
+### Storage Confirmation
+
+Confirm files written:
+- `.agentops/journal/ideas/{slug}/proposal.md` — the proposal document
+- `.agentops/journal/ideas/{slug}/scratchpad.md` — working notes
+- `.agentops/journal/ideas/{slug}/research.md` — research notes (Full development only)
+- `.agentops/journal/ideas/{slug}/exports/` — directory created for future exports
+
+### Index Management
+
+Update `.agentops/journal/ideas/index.json` using the atomic write protocol from conventions.md:
+1. Write updated JSON to `index.json.tmp`
+2. Read back and verify valid JSON
+3. Move to `index.json` via `mv`
+
+Add entry with: id (`idea-{slug}`), date, title, category, status (`Draft`), beneficiaries, auto-extracted tags, file path, cortex_id (null), notion_page_id (null).
+
+### Export
+
+Use `AskUserQuestion`: "Would you like to export the proposal?" — options: "PDF", "DOCX", "Both PDF and DOCX", "Skip export"
+
+If not skipped, follow the export approach from conventions.md:
+1. **Pandoc (primary):** `pandoc proposal.md -o {slug}-proposal.{ext}` — check availability via `which pandoc`
+2. **Playwright (fallback):** Render as styled HTML, then PDF
+3. **HTML (last resort):** Generate styled HTML with print instructions
+
+Save exports to `.agentops/journal/ideas/{slug}/exports/`.
+
+### Share Guidance
+
+Read contacts from `journal-config.json`. Match the idea's category (as snake_case, e.g., "Architecture improvement" maps to `architecture_improvement`) against each contact's `share_categories` array. Present matching contacts:
+
+> "Based on this idea's category, these contacts may be interested: {name} ({role}). You could share the exported proposal with them."
+
+Never send anything automatically. This is guidance only. If no contacts match or no contacts configured, skip silently.
+
+### Cortex Sync
+
+If Cortex is available and enabled in config: write the idea as a semantic memory via `mcp__cortex__cortex_write` with tags `["idea", "{category_snake_case}", "draft"]`. Store the returned ID as `cortex_id` in the index entry. On failure: log warning, continue.
+
+### Notion Sync
+
+If Notion is available and enabled in config:
+- If `ideas_database_id` is null: use `AskUserQuestion` — "Notion is connected but no ideas database exists. Create one?" — options: "Yes, create ideas database", "Skip Notion sync". If yes, create database with schema from conventions.md (Title, Category, Status, Date, Author, Beneficiaries, Summary) and store the ID in config.
+- Push the idea to the Notion database. Store `notion_page_id` in the index entry.
+- On failure: log warning, continue.
+
+### Auto-Commit
+
+If `preferences.auto_commit` is `true` in journal-config.json:
+
+1. Check whether `.agentops/journal/` is tracked by git (i.e., not excluded by `.gitignore`).
+2. If it is tracked, run:
+   ```
+   git add .agentops/journal/ideas/{slug}/
+   git commit -m "docs(idea): capture — {slug}"
+   ```
+3. If it is gitignored, skip auto-commit and warn: "Auto-commit skipped because `.agentops/journal/` is gitignored. To enable, unignore this path in `.gitignore`."
+
+If auto_commit is false (default), skip.
+
+---
+
+## Revisit Existing Idea
+
+When revisiting an existing idea (slug provided or selected):
+
+1. Read the existing proposal from `.agentops/journal/ideas/{slug}/proposal.md`
+2. Display current status and summary
+3. Use `AskUserQuestion`: "What would you like to do with this idea?" — options:
+   - "Update status"
+   - "Edit proposal content"
+   - "Export proposal"
+   - "View research notes"
+
+### Status Update
+
+Use `AskUserQuestion`: "Update status to:" — options: "Draft", "Proposed", "Approved", "Implemented", "Deferred"
+
+Record who updated the status (from `identity.name` in config) and when (UTC timestamp). Append a status history entry to the proposal:
+
+```
+### Status History
+| Date | Status | Updated By |
+|------|--------|-----------|
+| {date} | {new_status} | {identity.name} |
+```
+
+Update `.agentops/journal/ideas/index.json` with new status (atomic write). If Cortex available, update the memory tags. If Notion available, update the database entry. If auto_commit enabled:
+```
+git commit -am "docs(idea): update status — {slug} → {status}"
+```
+
+---
+
+## Error Handling
+
+- If conventions.md cannot be read, warn and use inline defaults
+- If config is malformed, offer to recreate via first-run setup
+- If index is corrupted, rebuild from proposal frontmatter per conventions.md rebuild protocol
+- If export tool is unavailable, fall through the export chain (Pandoc, Playwright, HTML)
+- Integration failures never block local operations — log and continue
+- Never leave partial writes — use atomic write protocol for all index updates

package/commands/enterprise/search.md

@@ -0,0 +1,236 @@
+---
+name: search
+description: Unified search across work log, ideas, git history, ADRs, lessons, and optional integrations
+---
+
+You are a unified search assistant for the work journal system. You search across all local and external knowledge sources and present results in a structured, actionable format.
+
+**Before starting, check the feature flag:**
+Run: `source hooks/feature-flags.sh && agentops_enterprise_enabled "work_journal"` — if disabled, inform the user: "Work journal commands are disabled. Enable with: /agentops:flags work_journal true" and stop.
+
+## CRITICAL RULE: Use AskUserQuestion Tool
+
+You MUST use the `AskUserQuestion` tool for EVERY question in this command. DO NOT print questions as plain text or numbered option lists. Call the AskUserQuestion tool which renders a proper selection UI. This is a BLOCKING REQUIREMENT.
+
+---
+
+## Startup
+
+1. **Read shared conventions:** Read `templates/work-journal/conventions.md` for config schema, storage layout, index format, and integration detection protocol.
+2. **Load config:** Read `.agentops/journal-config.json`. If missing, run the first-run setup defined in conventions.md.
+3. **Detect integrations** per the conventions Integration Detection protocol:
+   - Check Cortex: call `mcp__cortex__cortex_status`. Set CORTEX_AVAILABLE = true/false.
+   - Check Notion: check if `mcp__notion__notion_search` tool is available. Set NOTION_AVAILABLE = true/false.
+   - Check Linear: check if any Linear MCP tools are available. Set LINEAR_AVAILABLE = true/false.
+   - Report detected integrations once: "Integrations: Cortex [status], Notion [status], Linear [status]" or "Running in local-only mode."
+4. **Parse arguments:** The user's input is: $ARGUMENTS
+
+---
+
+## Mode Selection
+
+### Direct Search Mode (arguments provided)
+
+If `$ARGUMENTS` contains a query string (not just flags), search ALL available sources immediately with no prompts. Extract any `--since=YYYY-MM-DD` and `--until=YYYY-MM-DD` flags from arguments for date filtering. The remaining text is the search query.
+
+Proceed directly to **Execute Search** with scope = "Everything" and the parsed query.
+
+### Interactive Search Mode (no arguments)
+
+If `$ARGUMENTS` is empty or contains only flags:
+
+**Step 1 — Get query:** Call `AskUserQuestion`:
+- question: "What would you like to search for?"
+- header: "Search"
+- requireFreeText: true
+
+**Step 2 — Date range:** If no `--since`/`--until` flags were provided, call `AskUserQuestion`:
+- question: "Filter by time period?"
+- header: "Date Range"
+- options: [{label: "All time", description: "No date filter"}, {label: "This week", description: "Current week only"}, {label: "This month", description: "Current calendar month"}, {label: "Last 30 days", description: "Rolling 30 days"}, {label: "Last quarter", description: "Rolling 90 days"}, {label: "Custom range", description: "Specify start and end dates"}]
+
+If "Custom range" is selected, call `AskUserQuestion` for start date, then again for end date.
+
+**Step 3 — Scope selection:** Build scope options dynamically based on integration availability.
+
+Always include these options:
+- "Everything" — Search all available sources
+- "Work log only" — Search worklog entries
+- "Ideas only" — Search idea proposals
+- "Git history only" — Search commit history
+- "ADRs only" — Search architecture decision records
+- "Lessons & patterns only" — Search lessons and patterns
+
+Conditionally add:
+- If CORTEX_AVAILABLE: "Cortex memories" — Search cross-session memory
+- If LINEAR_AVAILABLE: "Linear issues" — Search project issues
+- If NOTION_AVAILABLE: "Notion pages" — Search workspace pages
+
+Call `AskUserQuestion`:
+- question: "Which sources should I search?"
+- header: "Search Scope"
+- options: [the dynamically built list above, each with a descriptive subtitle]
+
+---
+
+## Execute Search
+
+Search the selected sources. Where possible, run local and external searches in parallel (do not wait for one source before starting another).
+
+### Local Sources (always available)
+
+**(a) Work Log — worklog/index.json + file grep:**
+- Read `.agentops/journal/worklog/index.json`
+- Match entries where title, category, or tags contain the query keywords
+- If fewer than 5 index matches, run grep across `.agentops/journal/worklog/` markdown files for deeper matches
+- Apply date-range filter if specified
+- Limit to 20 results
+
+**(b) Ideas — ideas/index.json + file grep:**
+- Read `.agentops/journal/ideas/index.json`
+- Match entries where title, category, tags, or status contain the query keywords
+- If fewer than 5 index matches, run grep across `.agentops/journal/ideas/` markdown files for deeper matches
+- Apply date-range filter if specified
+- Limit to 20 results
+
+**(c) Git History:**
+- Run: `git log --all --grep="{query}" --max-count=20 --date=short --pretty=format:"%h %ad %s"`
+- If date-range specified, add `--since="{since}"` and `--until="{until}"`
+- Parse commit hash, date, and message from output
+
+**(d) Architecture Decision Records:**
+- Check if `docs/adr/` directory exists
+- If it exists, grep for the query across all files in `docs/adr/`
+- Extract title, date, and status from matching ADR files
+- Limit to 20 results
+
+**(e) Lessons & Patterns:**
+- Grep `tasks/lessons.md` for the query
+- Extract matching lesson/pattern sections with their titles and dates
+- Limit to 20 results
+
+### External Sources (conditional, searched in parallel with local)
+
+**(f) Cortex Memories (if CORTEX_AVAILABLE and scope includes Cortex):**
+- Call `mcp__cortex__cortex_search` with the query
+- Search across all memory types (semantic, episodic, procedural)
+- Limit to 20 results
+
+**(g) Linear Issues (if LINEAR_AVAILABLE and scope includes Linear):**
+- Use Linear MCP tools to search issues matching the query
+- Include issue ID, title, status, assignee, and date
+- Limit to 20 results
+
+**(h) Notion Pages (if NOTION_AVAILABLE and scope includes Notion):**
+- Call `mcp__notion__notion_search` with the query
+- Include page title, last edited date, and URL
+- Limit to 20 results
+
+---
+
+## Present Results
+
+Format results grouped by source. Only show sources that returned results. Skip empty sources silently.
+
+```markdown
+## Search Results for "{query}"
+
+**Filters:** {date range if applied, or "All time"} | **Scope:** {selected scope}
+**Sources searched:** {count} | **Total results:** {count}
+
+### Work Log ({count} results)
+| Date | Category | Entry | Path |
+|------|----------|-------|------|
+| {date} | {category} | {title/summary} | {file path} |
+
+### Ideas ({count} results)
+| Date | Category | Title | Status | Path |
+|------|----------|-------|--------|------|
+| {date} | {category} | {title} | {status} | {file path} |
+
+### Git History ({count} results)
+| Date | Commit | Message |
+|------|--------|---------|
+| {date} | {hash} | {message} |
+
+### Architecture Decisions ({count} results)
+| Date | ADR | Status | Path |
+|------|-----|--------|------|
+| {date} | {title} | {status} | {file path} |
+
+### Lessons & Patterns ({count} results)
+| Date | Type | Title | Path |
+|------|------|-------|------|
+| {date} | {pattern/lesson/anti-pattern} | {title} | {file path} |
+
+### Cortex Memories ({count} results)
+| Date | Type | Summary |
+|------|------|---------|
+| {date} | {semantic/episodic/procedural} | {summary} |
+
+### Linear Issues ({count} results)
+| ID | Status | Title | Assignee |
+|----|--------|-------|----------|
+| {id} | {status} | {title} | {assignee} |
+
+### Notion Pages ({count} results)
+| Last Edited | Title | URL |
+|-------------|-------|-----|
+| {date} | {title} | {url} |
+```
+
+If no results found across any source, inform the user: "No results found for '{query}'. Try broadening your search terms or adjusting the date range."
+
+---
+
+## Deep Dive
+
+After presenting results, build a list of the top results across all sources (up to 10 most relevant).
+
+Call `AskUserQuestion`:
+- question: "Want to dig deeper into any of these results?"
+- header: "Deep Dive"
+- options: [{label: "{source}: {title}", description: "{date} — {brief summary}"} for each top result, plus {label: "No, I have what I need", description: "End search"}]
+
+If the user selects a result:
+- Read and display the full content of the selected item
+- For local files: read the file and display the relevant section
+- For Cortex memories: call `mcp__cortex__cortex_read` with the memory ID
+- For Linear issues: fetch full issue details via Linear MCP tools
+- For Notion pages: fetch full page content via Notion MCP tools
+- After displaying, repeat the deep dive prompt with remaining results
+
+If "No, I have what I need" is selected, proceed to **Export**.
+
+---
+
+## Export
+
+Call `AskUserQuestion`:
+- question: "Would you like to export these search results?"
+- header: "Export"
+- options: [{label: "Export as markdown", description: "Save results to a markdown file"}, {label: "Export as PDF", description: "Generate a PDF report of results"}, {label: "No", description: "Skip export"}]
+
+### Export as Markdown
+- Write results to `.agentops/journal/exports/search-{query-slug}-{date}.md`
+- Include query, filters, and all results in the format above
+- Inform user of the file path
+
+### Export as PDF
+- Write markdown first, then use the export fallback chain from conventions.md:
+  1. Pandoc (primary): `pandoc input.md -o output.pdf`
+  2. Playwright (fallback): render as HTML, then PDF
+  3. HTML (last resort): generate styled HTML file
+- Save to `.agentops/journal/exports/search-{query-slug}-{date}.pdf`
+- Inform user of the file path
+
+---
+
+## Error Handling
+
+- If an index file is missing or corrupt, skip that source and note: "Worklog index not found — skipping. Run /agentops:worklog to create entries."
+- If an external integration fails mid-search, log a warning and continue with remaining sources. Never fail the entire search due to one source.
+- If git is not available, skip git history search silently.
+- If `docs/adr/` does not exist, skip ADR search silently.
+- If `tasks/lessons.md` does not exist, skip lessons search silently.
+- All searches are read-only. This command never modifies any files except when exporting results.