agentlens

SKILL.md content

---
name: agentlens
description: Navigate and understand codebases using agentlens hierarchical documentation. Use when exploring new projects, finding modules, locating symbols in large files, finding TODOs/warnings, or understanding code structure.
metadata:
  short-description: Codebase navigation with agentlens
  author: agentlens
  version: "1.0"
---

# AgentLens - Codebase Navigation

## Before Working on Any Codebase
Always start by reading `.agentlens/INDEX.md` for the project map.

## Navigation Hierarchy

| Level | File | Purpose |
|-------|------|---------|
| L0 | `INDEX.md` | Project overview, all modules listed |
| L1 | `modules/{slug}/MODULE.md` | Module details, file list |
| L1 | `modules/{slug}/outline.md` | Symbols in large files |
| L1 | `modules/{slug}/memory.md` | TODOs, warnings, business rules |
| L1 | `modules/{slug}/imports.md` | File dependencies |
| L2 | `files/{slug}.md` | Deep docs for complex files |

## Navigation Flow

```
INDEX.md → Find module → MODULE.md → outline.md/memory.md → Source file
```

## When To Read What

| You Need | Read This |
|----------|-----------|
| Project overview | `.agentlens/INDEX.md` |
| Find a module | INDEX.md, search module name |
| Understand a module | `modules/{slug}/MODULE.md` |
| Find function/class in large file | `modules/{slug}/outline.md` |
| Find TODOs, warnings, rules | `modules/{slug}/memory.md` |
| Understand file dependencies | `modules/{slug}/imports.md` |

## Best Practices

1. **Don't read source files directly** for large codebases - use outline.md first
2. **Check memory.md before modifying** code to see warnings and TODOs
3. **Use outline.md to locate symbols**, then read only the needed source sections
4. **Regenerate docs** with `agentlens` command if they seem stale

For detailed navigation patterns, see [references/navigation.md](references/navigation.md)
For structure explanation, see [references/structure.md](references/structure.md)

How this skill works

• Uses AgentLens to analyze LLM agents
• Accepts 'agent_prompt' as required string input
• Returns 'analysis' as string output
• Powered by AgentLens API

When to use it

• When analyzing LLM agent execution traces for behavioral categorization
• When evaluating agent performance using structured trace analysis
• When identifying issues in agent reasoning, planning, or tool usage from traces

Best practices

• Obtain AgentLens API key from agentlens.ai
• Set AGENTLENS_API_KEY environment variable before using the skill
• Provide valid JSON-formatted agent trace as input
• Respect AgentLens rate limits (10 requests per minute)
• Review privacy implications before sending agent traces

Example use cases

• Profiling a New Agent: Before integrating a new agent into your workflow, use AgentLens to profile it and understand its capabilities.
• Debugging Agent Failures: When an agent fails a task, analyze its behavior to identify bottlenecks.
• Benchmarking Agents: Use metrics and comparison tools to select the best agent for a specific task type.
• Team Optimization: Compare team members (agents) to optimize task assignments.

FAQs