Adrian's Work Modes — When to Call What

The human manual. Organized by how Adrian works, not by workflow internals. This doc resolves the naming collision: "workflows" = agent-side automated processes (the /slash commands). "Work modes" = how Adrian actually works.

Work Modes

🧠 Thinking Mode

"I want to think through something, no execution."

I want to...	Call	Notes
Discuss an idea openly	`/discuss`	No execution, no file edits
Score/compare options	`/discuss score`	Structured rubric comparison
Analyze tradeoffs	`/discuss analyze`	Deep pros/cons/tradeoffs
Get agent to probe my thinking	`/discuss probe`	Agent asks questions
Derive implications of a breakthrough	`/discuss derive`	推演 — 4-level derivation chain

Exit: Discussion saved in plans/ or session file. May trigger /task start if action-ready.

🔍 Research Mode

"I want to investigate something."

I want to...	Call	Notes
Research a topic	`/research`	Creates task, produces report
Evaluate a tool/tech	`/research evaluate`	Same + scoring matrix

Exit: Report published to portal. May trigger /task start to implement findings.

💻 Coding Mode

"I'm building something."

I want to...	Call	Notes
Start a task	`/task start`	Creates folder, ongoing entry, D1 record
Checkpoint my work	`/task save`	Commit + log. Task stays active
Mark task complete	`/task done`	Updates ongoing + D1. Doesn't commit/log (wrapup does that)
Spawn future work	`/task spawn`	Full work package for another session
Delegate quick work	`/task delegate`	Light handoff instructions for another session

📋 Session Lifecycle

"Managing my work session."

I want to...	Call	Notes
Check what I'm working on	`/hub-status`	Read-only pulse, 15 lines
End my session	`/wrapup`	Orchestrates: learn → commit → log → evolve
End my day	`/wrapup day`	Wraps ALL tasks + daily summary
Pass to fresh session (context bad)	`/handoff` → `/wrapup`	Save state first, then close

🧭 Alignment Mode

"Am I doing the right thing?"

I want to...	Call	Notes
Check vision alignment	`/vision` or `/vision align`	Am I serving the vision?
Find missing purpose	`/vision check`	Are there purposeless items?
Audit workspace for drift	`/vision scan`	Full workspace vision audit
Measure clarity	`/clarity`	Crystallization assessment
Full clarity report	`/clarity report`	Breakdown + next steps

🔄 Hub Maintenance Mode

"The system itself needs care."

I want to...	Call	Notes
Sync docs after changes	`/hub-evolve`	After modifying workflows/rules/skills
Deep quality inspection	`/hub-audit`	Scan against principles
Data hygiene	`/hub-clean`	Log integrity, task consistency
Archive done tasks	`/hub-archive`	Move to archive folders

The Session Flow (typical day)

Morning:
  /hub-status                ← "where am I?"
  /task start TASK_ID        ← pick up work

Working:
  (code/discuss/research)
  /task save                 ← checkpoint if needed

Insight hits:
  /discuss derive            ← derive implications

End of session:
  /wrapup                    ← close everything out
    └── /learn               ← extract lessons (auto)
    └── /commit              ← commit changes (auto)
    └── /log                 ← log entry (auto)
    └── /hub-evolve          ← if hub changed (auto)

End of day:
  /wrapup day                ← wrap all + daily summary

Quick Reference

Think:    /discuss [score|analyze|probe|derive]
Research: /research [evaluate]
Build:    /task [start|save|done|spawn|delegate]
Align:    /vision | /clarity
Close:    /wrapup [day]
Maintain: /hub-evolve | /hub-audit | /hub-clean | /hub-archive
Pulse:    /hub-status

Part of the Cortex R1B-2 review. This is the human manual — organized by how Adrian works, not by agent internals.

Work Modes

On this page