darwin_skill

This skill orchestrates project ecosystem evolution by sensing signals, evaluating agent fitness, and proposing coordinated improvements to boost reliability
  • Shell

8

GitHub Stars

1

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill simota/agent-skills --skill darwin

  • SKILL.md8.4 KB

Overview

This skill is an ecosystem self-evolution orchestrator that senses project state, assesses agent fitness, and proposes targeted evolution actions. It integrates existing scores and agent journals to compute an Ecosystem Fitness Score (EFS) and per-agent Relevance Scores (RS). The skill aims to guide incremental improvements, surface sunset candidates, and persist ecosystem intelligence for continuous adaptation.

How this skill works

Darwin collects signals from git metrics, file structure, activity logs, agent journals, and existing scores to detect the project lifecycle and compute EFS. It evaluates each agent’s Relevance Score, runs trigger checks (ET-01…ET-08), and proposes evolution actions like affinity overrides, journal synthesis, discovery propagation, and sunset recommendations. All observations and actions are written to .agents/ECOSYSTEM.md for verification and trend tracking.

When to use it

  • You need a holistic health check of an AI agent ecosystem
  • You want prioritized proposals to improve agent collaboration or coverage
  • You suspect agents are stale or should be retired
  • A lifecycle phase shift may require routing or priority changes
  • You need cross-agent patterns extracted from journals

Best practices

  • Run full cycle regularly and after major repo changes to keep EFS current
  • Prefer proposals over automatic changes; require Void confirmation before sunset
  • Correlate RS changes with actual invocation data before acting
  • Use confidence thresholds (≥0.60) to report single-phase lifecycle vs mixed state
  • Limit simultaneous affinity overrides and ask stakeholders for large changes

Example use cases

  • Run /Darwin to produce a DARWIN_REPORT showing phase, EFS, and evolution proposals
  • Invoke /Darwin lifecycle after a release to confirm phase and trigger affinity updates
  • Use /Darwin journals to synthesize reusable patterns from agent notes into Pattern Cards
  • Trigger /Darwin staleness to identify dormant agents and flag sunset candidates for Void verification
  • Run /Darwin fitness to produce a dimensioned EFS dashboard for stakeholder review

FAQ

No. Darwin proposes changes and writes rationale to ECOSYSTEM.md; it never edits agent definitions or take irreversible actions.

How does Darwin decide to recommend sunset?

It flags agents with RS <20 and corroborates with lifecycle context and usage trends, then asks Void for YAGNI verification before retirement.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational
darwin skill by simota/agent-skills | VeilStrat