Can I run this for a single agent instead of a whole fleet?

Yes, the skill includes a Single-Agent Analysis mode that runs the capability check and gap analysis inline without spawning the entire agent fleet.

How does the gap analysis categorize findings?

Findings are sorted into three categories: Underclaimed (in description but not reported), Undocumented (reported but not in description), and Scope Violations (capabilities outside the agent's intended domain).

Does this skill require external databases?

It uses a local LevelDB store (agent-map.db) relative to your project directory, requiring the 'level' npm package to be installed.

What is description drift in AI agents?

Description drift occurs when an agent's frontmatter description no longer matches what the agent actually does, causing orchestrators to make incorrect routing decisions.

Agent Capability Analyzer

Name: Agent Capability Analyzer
Author: Jamie-BitFlight

byJamie-BitFlight

•

분석 및 모니터링

Quantifies description drift in AI agents by comparing self-reported capabilities against static frontmatter to ensure accurate orchestrator routing.

The Agent Capability Analyzer is a diagnostic tool designed for multi-agent systems to solve the problem of 'description drift'—where an agent's metadata no longer matches its actual behavior. By spawning agents to self-report their capabilities and comparing these responses against static frontmatter descriptions, it provides a structured gap analysis. This tool is essential for developers building complex AI workflows who need to audit agent reliability, maintain a LevelDB-backed agent map, and ensure that orchestrators are making routing decisions based on accurate, up-to-date information.

주요 기능

01Automated description drift detection and gap analysis

0230 GitHub stars

03Standardized prompt templates for consistent capability collection

04LevelDB-backed storage for persistent agent capability mapping

05Categorized reporting of underclaimed, undocumented, and scope-violating behaviors

06Supports both simultaneous multi-agent fleet audits and single-agent checks

사용 사례

01Detecting undocumented agent capabilities that the orchestrator is unaware of

02Auditing a fleet of specialized agents to identify misrouting risks

03Validating that specialized agents are staying within their intended functional scope

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jamie-bitflight/claude_skills agent-capability-analyzer

For use in Claude.ai and ChatGPT

주요 기능

01Automated description drift detection and gap analysis

0230 GitHub stars

03Standardized prompt templates for consistent capability collection

04LevelDB-backed storage for persistent agent capability mapping

05Categorized reporting of underclaimed, undocumented, and scope-violating behaviors

06Supports both simultaneous multi-agent fleet audits and single-agent checks

사용 사례

01Detecting undocumented agent capabilities that the orchestrator is unaware of

02Auditing a fleet of specialized agents to identify misrouting risks

03Validating that specialized agents are staying within their intended functional scope

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jamie-bitflight/claude_skills agent-capability-analyzer

For use in Claude.ai and ChatGPT