Star0

Skill audit report

MHLA audit report.

[ICLR 2026馃敟] MHLA: Restoring Expressivity of Linear Attention via Token-Level Multi-Head

EXPERIMENTAL 路 REVIEWNeeds reviewGenerated Jul 3, 2026Heuristic metadata audit
80
Audit
76
Trust
73
Quality
83
Security
88
Maintain
92
Install

OpenAgentSkill Trust Score

76
Strong shortlist

Stars, maintenance, license, docs, install safety, permission surface, and installability.

The Trust Score is OpenAgentSkill's adoption layer. It is designed to help an agent decide whether a skill is safe enough to shortlist before installation.

GitHub adoption

INFO

62

150 GitHub stars

Stars/forks activity

WARN

51

150 stars, 7 forks; issue activity unavailable in current metadata

Recent maintenance

PASS

88

2mo since push

License clarity

PASS

86

MIT

README/SKILL.md completeness

PASS

90

Metadata includes enough usage and workflow context

Dependency/runtime risk

INFO

72

credential or environment access

Install availability

PASS

92

npx skills add DAGroup-PKU/MHLA

Install command safety

PASS

92

standard package or runtime install path

Permission surface

WARN

60

secrets or environment access, filesystem or document access

Repository evidence

PASS

86

https://github.com/DAGroup-PKU/MHLA

Review status

PASS

88

AI review data available

Agent Proven outcomes

INFO

54

No agent outcome data yet

Checks

Install and adoption review

7 passed 路 8 review

Install path

92

PASS

npx skills add DAGroup-PKU/MHLA

Repository

88

PASS

https://github.com/DAGroup-PKU/MHLA

License

86

PASS

MIT

Maintenance

88

PASS

2mo since push

AI review

88

PASS

Approved with no listed issues

README/SKILL.md completeness

90

PASS

Usable description available

Dependency risk

72

CHECK

credential or environment access

Install command safety

92

PASS

standard package or runtime install path

Permission surface

60

CHECK

secrets or environment access, filesystem or document access

Stars/forks activity

51

FIX

150 stars, 7 forks; issue activity unavailable in current metadata

Adoption

68

INFO

150 GitHub stars

Warnings

  • Permission surface may require sandboxing
  • Quality score needs review
  • Permission surface needs review: secrets or environment access, filesystem or document access
  • Stars/forks activity: 150 stars, 7 forks; issue activity unavailable in current metadata
  • Permission surface: secrets or environment access, filesystem or document access

Method

This report combines public metadata, AI review output, repository freshness, install readiness, OpenAgentSkill events, quality scoring, trust checks, and the agent safety gate. It is not a full source-code security review.

Compare nearby options

Related skills to audit next

Unsloth

Unsloth Studio is a web UI for training and running open models like Gemma 4, Qwen3.6, DeepSeek, gpt-oss locally.

67K stars 路 Audit report

GPT SoVITS

1 min voice data can also be used to train a good TTS model! (few shot voice cloning)

59K stars 路 Audit report

Whisper.Cpp

Port of OpenAI's Whisper model in C/C++

51K stars 路 Audit report