Does it check if an agent has the right tools for a specific prompt instruction?

Yes, one of the primary audit dimensions is Tool Access Sufficiency, which maps prompt instructions to required tools and verifies their availability in the agent configuration.

Where are the audit reports saved?

All audit artifacts, including markdown reports and dependency graphs, are written to the .claude/audits/ directory within your specified plugin path.

What is the difference between skill-lifecycle-audit and agent-lifecycle-audit?

Skill-lifecycle-audit focuses on workflow coherence across SKILL.md files, while agent-lifecycle-audit focuses on the actual execution capability of agent files, checking their tool access and prompt alignment.

Can this skill automatically fix the issues it finds?

Yes, it can flag significant 'description drift' or scope violations and pass them to the subagent-refactorer for automated remediation and prompt optimization.

What does the 'dead agent' detection feature do?

It identifies agent files that are never referenced by skills, commands, or other agents, helping you clean up redundant code and optimize plugin maintenance.

Agent Lifecycle Auditor

Name: Agent Lifecycle Auditor
Author: Jamie-BitFlight

byJamie-BitFlight

•

セキュリティとテスト

Validates agent execution capabilities by auditing the alignment between configurations, prompts, and tool access.

The Agent Lifecycle Auditor is a sophisticated diagnostic tool designed to ensure AI agents within a Claude Code plugin can actually perform their documented tasks. It performs eight distinct semantic audits—ranging from tool sufficiency checks to inter-agent contract validation—to identify execution gaps that static analysis might miss. By detecting 'description drift,' contradictory prompt instructions, and 'dead agents,' this skill helps developers maintain high-quality, reliable agentic workflows and ensures that delegated tasks are handled by agents with the correct resources and permissions.

主な機能

01Automated detection of 'description drift' between frontmatter and self-reported capabilities

028-dimension semantic audit of agent configurations and execution capabilities

0326 GitHub stars

04Generation of comprehensive audit reports and visual agent-to-skill dependency graphs

05Inter-agent delegation contract validation to ensure seamless task handoffs

06Identification of contradictory prompt instructions and missing tool permissions

ユースケース

01Post-refactoring validation to confirm that agent-to-agent delegation logic remains intact

02Troubleshooting complex agentic workflows where sub-agents are failing to execute tasks

03Pre-marketplace review to ensure agent quality and documentation accuracy

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jamie-bitflight/claude_skills audit-agent-lifecycle

For use in Claude.ai and ChatGPT

Download Skill