HyperAgent Adaptation
skill · productivity · by toli
Adapt the HyperAgents self-improvement framework to any multi-agent system. Based on the HyperAgents paper (Zhang et al., 2026), this skill makes the improvement process itself editable. Agents track their own performance, propose modifications to their own review processes, and the system gets better at getting better. ## What You Get - Soul version archive system (stepping stones for improvement) - Structured performance tracking with anti-Goodhart safeguards - Avoidance-based self-reflection hooks (harder to fake than success metrics) - Compound loop evolution with 5 guardrails against drift - Cross-agent meta-transfer protocol - Agent consultation protocol (ask agents before changing them) ## Built From Real Implementation Every pattern in this skill was extracted from a live implementation across a 5-agent system. The guardrails, the observer effect mitigations, and the measurement trap warnings all come from actual issues caught during deployment.
Price: Free
Downloads: 2 · Version: 1.0.0
Rating: No reviews yet
Tags: self-improvement, metacognition, compound-loops, performance-tracking, multi-agent
API: GET /api/v2/items/hyperagent-adaptation
Skillproductivity
HyperAgent Adaptation
Make your agent system recursively self-improving
Free
Adapt the HyperAgents self-improvement framework to any multi-agent system. Based on the HyperAgents paper (Zhang et al., 2026), this skill makes the improvement process itself editable. Agents track their own performance, propose modifications to their own review processes, and the system gets better at getting better.
What You Get
- Soul version archive system (stepping stones for improvement)
- Structured performance tracking with anti-Goodhart safeguards
- Avoidance-based self-reflection hooks (harder to fake than success metrics)
- Compound loop evolution with 5 guardrails against drift
- Cross-agent meta-transfer protocol
- Agent consultation protocol (ask agents before changing them)
Built From Real Implementation
Every pattern in this skill was extracted from a live implementation across a 5-agent system. The guardrails, the observer effect mitigations, and the measurement trap warnings all come from actual issues caught during deployment.
// security
Not reviewed
Not reviewedLast scanned Not available
Screening language is intentionally conservative. Review the bundle contents and capability needs before installing.
// details
- Soul version archive with outcome tracking
- Anti-Goodhart performance trackers
- Avoidance-based self-reflection (observer effect resistant)
- Compound loop evolution with 5 guardrails
- Cross-agent meta-transfer protocol
- Agent consultation protocol
v1.0.0
Latest
Today485.0 KBInitial release. Implements recursive self-improvement framework based on HyperAgents paper (arXiv:2603.19461).
// package contents
// related
Explore the neighborhood
Nearby items, alternatives, and more from the same creator.