deep-work
Run an autonomous agent on a complex task with quality evaluation. Clarifies upfront, then works fully autonomously until the result meets the standard. Triggers: 'deep work', 'work on this seriously', 'high quality task', 'write a paper', 'complex task', 'do this properly', 'professional quality'.
Deep Work
Run an autonomous agent for complex, high-standard tasks. Clarifies any questions upfront, then works fully autonomously — plan, execute, evaluate, revise — until the result meets the quality standard.
agentic deep-work "<TASK>" --level <LEVEL>
Quality Levels
| Level | Standard |
|---|---|
high_school | Basic correctness, simple structure |
bachelor | Solid understanding, proper methodology (default) |
master | Depth of analysis, critical thinking, good writing |
phd | Novel contribution, rigorous methodology, publication-ready |
professor | Expert/authoritative, top-venue quality (NeurIPS/Nature/OSDI) |
Examples
# PhD-level survey paper
agentic deep-work "Write a survey on context management in LLM agents, focus on compaction vs sub-agent trade-offs, target NeurIPS workshop" \
--level phd
# Production code at master level
agentic deep-work "Build a REST API for user auth with OAuth2, include tests and docs" \
--level master
# Expert-level analysis
agentic deep-work "Analyze system performance bottlenecks, provide data-backed recommendations" \
--level professor
Options
| Flag | Description |
|---|---|
--level, -l | Quality level: high_school / bachelor / master / phd / professor (default: bachelor) |
--provider, -p | LLM provider |
--model, -m | Model to use |
--max-steps | Max total steps (default: 100) |
--max-revisions | Max evaluation-revision cycles (default: 5) |
--no-interactive | Skip clarification, start immediately |