Menu
Deep research protocol — ALWAYS load this skill before any research task. For multi-stream research with parallel domain agents, use research-coordinator instead.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Cold email outreach agent for plusvibe.ai. Campaign planning, subject line optimization, audience segmentation, drip sequence design, deliverability strategy. Writes all outputs to MemOS email-mkt-cube.
Master research orchestration skill. Decomposes any research query into parallel domain streams, spawns specialized researcher agents, and synthesizes results into a comprehensive intelligence brief. Replaces deep-research for multi-stream research tasks. Use Opus-class model for this role.
Domain researcher for academic and intellectual discourse — orchestrates arxiv and hn-research to cover cutting-edge papers, citation landscapes, and technical community reaction on Hacker News. Returns a structured academic intelligence report.
Implement Karpathy's autoresearch self-improving agent loop framework. Use when researching autonomous AI research agents, self-optimizing systems, or integrating test-time compute scaling into agent workflows.
Domain researcher for code and ML ecosystem signals — orchestrates github-research and huggingface-hub to produce a unified code intelligence report covering repos, releases, models, datasets, and community activity.
Research GitHub repositories, topics, trending projects, issues, discussions, and community activity. Use when user asks about GitHub projects, open-source tools, repo stats, contributor activity, or any task requiring GitHub data.
| name | deep-research |
| description | Deep research protocol — ALWAYS load this skill before any research task. For multi-stream research with parallel domain agents, use research-coordinator instead. |
For comprehensive multi-stream research, load
research-coordinatorinstead. It orchestrates parallel domain researchers (social media, code/ML, academic, market intelligence) and synthesizes a full intelligence brief.
This skill covers the simpler case: single-agent research with direct skill access.
MUST LOAD THIS SKILL BEFORE ANY RESEARCH TASK.
| # | Skill | Purpose |
|---|---|---|
| 1 | web-research | Web search, scraping, extraction |
| 2 | youtube-content | YouTube transcripts and video research |
| 3 | xitter | X/Twitter discussions and sentiment |
| 4 | github-research | GitHub repos, stars, issues, activity |
| 5 | reddit-research | Reddit communities and discussions |
STEP 1: skill_view() each of the 5 skills above IN ORDER
(web-research, youtube-content, xitter, github-research, reddit-research)
STEP 2: For each skill, note:
- Required tools/commands
- Rate limits or usage notes
- Key gh commands (for github-research)
STEP 3: Plan research subagents
- Assign specific skills to each subagent based on their research focus
- Pass the relevant skill names in the skills parameter
STEP 4: Spawn subagents with skills=[...] attached
Example:
sessions_spawn(
task: "Research X using web-research + github-research + xitter skills",
skills: ["web-research", "github-research", "xitter"]
)
STEP 5: Synthesize results from all subagents
❌ DO NOT spawn subagents with just toolsets: ["web"]
✅ ALWAYS load the 5 skills first, then pass skills to subagents
❌ DO NOT skip skill loading "to save time" ✅ Research quality depends on having the right tools available
❌ DO NOT skip xitter if the topic likely has X/Twitter discussion ✅ Social sentiment is part of comprehensive research
Automatically load this skill when user says:
If a research task is requested and this skill was NOT loaded, go back and load it before proceeding. No exceptions.