Menu
Run parallelized deep research across multiple angles. Generate specs, spawn Claude agents in tmux, monitor progress, fix incomplete reports, and compile a styled PDF.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Smash a GitHub issue end-to-end with Codex: create a feature branch, implement the fix, commit, open a PR, request deep review, wait, and then either fix P1 findings or post a success comment. Use when a user wants a no-nonsense issue execution flow from a GitHub issue URL.
Interact with Kaizen OS via MCP API to manage cards, workitems, scratchpad notes, and calendar events. Use when tasks involve planning, personal productivity, or scheduling.
Generate or edit images via Gemini 3 Pro Image (Nano Banana Pro).
Search the web with Tavily API via curl. Use when users explicitly request Tavily, when you need Tavily-specific controls (search_depth, topic, time_range, include/exclude domains, include_answer, include_raw_content), or when built-in web_search provider setup is unavailable.
Query and update Twenty CRM person records via REST. Use when a task requires finding a person by name, pulling full person data, or updating a person record in Twenty.
| name | wide-research |
| description | Run parallelized deep research across multiple angles. Generate specs, spawn Claude agents in tmux, monitor progress, fix incomplete reports, and compile a styled PDF. |
| metadata | {"openclaw":{"requires":{"bins":["claude","tmux","jq","python3"],"env":["REPORT_PATH"]},"primaryEnv":"REPORT_PATH"}} |
Parallelized deep-research skill. Takes a topic, generates research angles, spawns Claude agents in tmux sessions with a worker pool, and compiles results into a styled PDF.
Set REPORT_PATH in environment or in {baseDir}/.env:
export REPORT_PATH="$HOME/.ocreports"
Default is ~/.ocreports if not set.
Claude generates a spec JSON and pipes it to init.sh:
cat <<'EOF' | {baseDir}/scripts/init.sh my-research
{
"description": "AI agent frameworks comparison",
"angles": [
{ "label": "01", "title": "LangChain Analysis", "task": "Research LangChain architecture..." },
{ "label": "02", "title": "CrewAI Analysis", "task": "Research CrewAI patterns..." }
]
}
EOF
This creates the project directory with stubs and specs.json.
{baseDir}/scripts/run.sh ~/.ocreports/my-research
Options:
--fix-only Only re-run incomplete reports
--dry-run Show what would happen without launching
--max-workers N Max concurrent tmux sessions (default: 10)
{baseDir}/scripts/status.sh ~/.ocreports/my-research
{baseDir}/scripts/run.sh ~/.ocreports/my-research --fix-only
{baseDir}/scripts/report.sh ~/.ocreports/my-research
Options:
--output <filename> Override output filename
Requires Python packages: markdown, weasyprint. The script auto-installs them if missing.
{baseDir}/scripts/cleanup.sh ~/.ocreports/my-research
Use --force to skip confirmation.
{
"description": "Project description for the cover page",
"angles": [
{
"label": "01",
"title": "Human-readable title",
"task": "Detailed research instructions for the agent"
}
]
}
| Variable | Default | Description |
|---|---|---|
REPORT_PATH | ~/.ocreports | Base directory for all research projects |
MAX_WORKERS | 10 | Max concurrent tmux agent sessions |
POLL_INTERVAL | 10 | Seconds between worker pool checks |
MODEL | (none) | Claude model override |