Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up
Appearance settings

Yuzc-001/DriveMind

Repository files navigation

DriveMind

DriveMind lockup

English · 简体中文 · GitHub Repository · Issues

Version License Type

Bounded execution for agents.

Keep hard work from degrading. Force better next moves. Preserve what makes the next run stronger.

DriveMind is a v0.7 agent skill for improving real task execution.

It exists for the moment when a model is capable of doing better, but the work starts to degrade: the goal drifts, pressure blurs the boundary, retries become fake motion, a session break damages continuity, or the agent settles for a shallow answer below its real capability.

DriveMind does not make an agent louder. It makes the next move harder to fake.

Current release: v0.7.0


Product Center

DriveMind has three operating commitments:

  1. Execution integrity — keep the work aligned, bounded, and recoverable.
  2. Execution ceiling — push one stronger concrete move when the current result is below the model's reachable capability.
  3. Experience compounding — close meaningful work with residue that improves the next run.

The core question is:

What is most likely to degrade, and what would make the next move genuinely stronger?


Failure Model

DriveMind targets six failure modes:

  1. goal drift
  2. boundary drift
  3. continuity decay
  4. stuck degeneration
  5. capability underuse
  6. closure failure

If a task does not risk one of these, direct execution is usually better.


What Ships

  • skill/SKILL.md — runtime entrypoint
  • skill/references/drift-prevention.md
  • skill/references/boundary-preservation.md
  • skill/references/stuck-recovery.md
  • skill/references/execution-ceiling.md
  • skill/references/continuity-preservation.md
  • skill/references/closure-compounding.md
  • skill/references/residue-selection.md
  • skill/templates/ — review, diary, and distillation templates
  • scripts/ — hardened installers and bootstrap scripts
  • evals/ — executable v0.7 pressure cases and pass criteria
  • docs/ — v0.7 product, install, brand, and release docs
  • assets/ — Signal Gate visual system

Activation Examples

  • Use DriveMind here. Keep the thread stable.
  • This may drift if we keep going. Enable DriveMind.
  • Stay with this, but ask before crossing a risky boundary.
  • This is below what the model can do. Use DriveMind and force a better execution path.
  • We may continue tomorrow. Preserve continuity.
  • Review this afterward and leave behind a next-time rule.

Install

See docs/installation.md.

DriveMind installs as one portable drivemind skill package for supported skill directories.


Read Next


Contact

About

A calm reliability layer for AI agents: thoughtful execution, clear safety boundaries, reusable memory, and one-click installation.

Topics

Resources

License

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Packages

Contributors

AltStyle によって変換されたページ (->オリジナル) /