OpenAgentSkill
GitHub200+

Pre-install eval

ai-agent-builder Eval report.

A machine-readable install decision for agents: task fit, Trust Score, Audit Score, install safety, permission surface, and a concrete validation plan before this skill touches a workspace.

Needs reviewMedium riskREVIEW Policy
63
Eval report
65
Trust
67
Audit
47
Agent safety

manual review

Test manually in an isolated workspace and compare against safer alternatives.

Required gates

Checks an agent must pass before install

Open JSON

Task fit

94

pass

Task wording matches this skill metadata.

  • Evaluate ai-agent-builder before installing it in an agent workflow
  • other
  • Workflow automation workflows; Claude Code teams; builders willing to evaluate younger projects

Install path

92

pass

Install handoff is available.

  • npx skills add claude-office-skills/skills --skill ai-agent-builder

Install command safety

92

pass

standard package or runtime install path

  • npx skills add claude-office-skills/skills --skill ai-agent-builder

Trust score

65

warn

Potentially useful, but at least one trust signal needs human inspection.

  • Manual review
  • 498 GitHub stars
  • MIT

Audit score

67

warn

Needs review

  • Permission surface may require sandboxing

Agent safety gate

47

warn

Sparse or mixed signals. Useful for discovery, but not for autonomous installation.

  • Test manually in an isolated workspace and compare against safer alternatives.
  • Permission surface may require sandboxing

License clarity

86

pass

MIT

  • MIT

Permission surface

60

warn

filesystem or document access, network or browser access

  • Network access: medium
  • Filesystem access: medium
  • Database access: medium

Validation plan

What the agent should do next

  1. 1Inspect repository, README/SKILL.md, license, and recent commits before production use.
  2. 2Install in an isolated workspace or sandbox with no production secrets available.
  3. 3Run the smallest representative task and record files touched, commands run, network access, and outputs.
  4. 4Compare the selected skill against at least one alternative when the eval status is review or failed.
  5. 5Promote only after the agent reports a successful verification result and unresolved warnings are accepted.

Do not use when

Conditions that require another skill

  • teams that need a vendor-supported SLA
  • production agents without a repository review
  • SKILL.md does not include an explicit setup/prerequisites section, so it is unclear what tools, APIs, keys, or runtimes are required to apply the patterns.
  • No OpenAgentSkill engagement data yet
  • Permission surface may require sandboxing
  • No limitations or safe operating boundaries are defined, such as when not to use autonomous agents or which actions require human approval.

Supporting checks

Trust signals behind the decision

README/SKILL.md completeness

pass

86

Metadata includes enough usage and workflow context

Recent maintenance

warn

62

9mo since push

Alternatives available

pass

82

Alternative skills are available for comparison.