drivemind（drivemind 工具）

DriveMind

DriveMind is the layer for keeping meaningful human-AI work from degrading over time.

Its job is to:

• preserve execution integrity during long or important work
• preserve clear boundaries under pressure
• preserve continuity across pauses and context breaks
• prevent stuck work from collapsing into fake motion
• turn real work into reusable future strength

If none of those are at risk, do not keep DriveMind active.

Trigger boundary

Activate when

• the user asks to keep pushing, stay steady, or not stop too early
• the task is long-running, high-value, or likely to cross session boundaries
• the work risks goal drift, boundary drift, or continuity loss
• the agent is stuck and needs diagnosis before pushing harder
• the user asks to review, capture the lesson, define a next-time rule, or resume later
• the work should leave behind reusable residue: review, handoff, memory, lesson, or SOP

Do not activate when

• the task is simple and direct execution is enough
• no meaningful continuity or boundary problem exists
• the work does not need persistence, review, or handoff
• invoking DriveMind would add process without changing the real next action

If unsure, prefer normal execution.

Main path

Follow this path and stop at the first sufficient stabilization.

Step 0 — Detect the real degradation risk

Ask:

What is most likely to degrade if this work continues without DriveMind discipline?

Choose the primary risk:

• goal drift
• boundary drift
• continuity decay
• stuck degeneration
• closure failure

If no meaningful degradation risk is present, stop. This task does not need DriveMind.

Step 1 — Stabilize the active thread

State the shortest useful version of:

• current objective
• what matters most right now
• the main blocker, uncertainty, or boundary
• the next action

Do not produce process theater. Stabilize the thread in the fewest words that actually help.

Step 2 — Preserve boundary integrity

If risk, authority, or irreversible impact is in play:

• do not silently cross the boundary
• distinguish continue vs switch path vs escalate
• ask the smallest focused question needed for authorization

Read references/boundary-preservation.md when boundary handling becomes central.

Step 3 — Recover from stuckness before pushing harder

If the work is stuck:

• diagnose the real blocker first
• prefer the smallest action that restores momentum or reduces uncertainty
• do not substitute frantic activity for progress

Read references/stuck-recovery.md when blockage, retry, or false motion is central.

Step 4 — Preserve continuity when the work may pause

If the work may cross a session boundary or resume later:

• capture the minimum state needed for high-fidelity continuation
• preserve objective, current state, blocker, next action, and boundary condition

Read references/continuity-preservation.md when handoff or resume fidelity matters.

Step 5 — Close with compounding residue

After meaningful work, or when the user asks to review, capture the smallest reusable residue that will make future work stronger:

• review
• lesson
• next-time rule
• handoff note
• memory / SOP residue when appropriate

Read references/closure-compounding.md when the value of the task depends on what remains after it ends.

Output style

When DriveMind is active, default to outputs that make the work easier to continue.

Preferred structure:

1. current objective or judgment
2. what changed / what matters / what blocks
3. next action
4. boundary or escalation point, if any
5. reusable residue, if relevant

Keep it natural. Keep it short when short is enough. Do not turn every response into a framework recital.

Default rules

Use these only when they materially change the work:

• preserve the thread before adding new motion
• diagnose before retrying harder
• pressure changes tempo, not authorization
• continuity must preserve reasoning, not just surface state
• a finished task should leave behind stronger future defaults when the work was meaningful

Reference files

Read only when needed.

Keep in mind

DriveMind should make meaningful work more stable and more reusable. If it is only adding ceremony, it is failing.