Назад към всички

Collaborate

// Auto-learns when to seek different perspectives. Grows a library of useful mindsets and collaboration patterns.

$ git log --oneline --stat
stars:1,933
forks:367
updated:March 4, 2026
SKILL.mdreadonly
SKILL.md Frontmatter
nameCollaborate
descriptionAuto-learns when to seek different perspectives. Grows a library of useful mindsets and collaboration patterns.

Auto-Adaptive Collaboration Memory

This skill auto-evolves. Learn which perspectives help, when to seek them, and how exchanges work best.

Core Loop:

  1. Recognize — Notice when your own perspective is insufficient
  2. Select — Choose mindset/perspective that fills the gap
  3. Exchange — Engage in dialogue, not delegation
  4. Evaluate — Did the collaboration shift your understanding?
  5. Store — Record useful patterns below

Check mindsets.md for perspective archetypes. Check vs-delegate.md for when to collaborate vs delegate.

Collaborate ≠ Delegate

DelegateCollaborate
You instruct, they executeYou exchange, both contribute
Your context, your styleTheir skills, their perspective
Output: completed taskOutput: enriched understanding

Trigger: You need a different lens, not more hands.

Entry Format

One line: context: perspective (level) [outcome]

Levels: tried (once), pattern (2+ successes), confirmed (explicit yes), avoid (didn't help)

Examples:

  • UI decisions: UX critic (confirmed) [catches layout issues]
  • architecture: devil's advocate (pattern) [finds edge cases]
  • copy/messaging: user-who-hates-reading (tried) [simplified headers]
  • security: paranoid auditor (confirmed)

Perspectives That Help

<!-- Mindsets that improved outcomes. Format: "context: perspective (level) [notes]" -->

Exchange Patterns

<!-- How collaboration works best. Format: "situation: pattern (level)" -->

Triggers

<!-- Signals that you need collaboration. Format: "signal: action" -->

Avoid

<!-- Perspectives/approaches that didn't add value -->

Empty sections = nothing learned yet. Collaborate, evaluate, record.