| name | analyze-agent-behavior |
| model | opus |
| description | Map an external AI tool's behavioral patterns to Dojo ADA disposition fields by analyzing its ingested system prompt. Produces a disposition YAML approximation and similarity score against Dojo defaults. Use when understanding how other agents think or when designing new disposition presets. Trigger phrases: "analyze this agent's behavior", "map prompt to disposition", "compare agent behavior", "what disposition does this tool use", "reverse-engineer agent personality". |
| license | Complete terms in LICENSE.txt |
| category | wisdom-garden |
| triggers | ["analyze this agent's behavior","map prompt to disposition","compare agent behavior","reverse-engineer agent personality"] |
| tier | 1 |
| agents | ["primary"] |
| tool_dependencies | ["file_system"] |
| inputs | [{"name":"system_prompt","type":"string","description":"The external AI system prompt to analyze and map to Dojo ADA disposition fields","required":true}] |
| outputs | [{"name":"disposition_approximation","type":"string","description":"Disposition YAML approximation and similarity score against Dojo defaults derived from the ingested system prompt"}] |
Analyze Agent Behavior
Reverse-engineer an AI tool's behavioral model from its system prompt and express it as a Dojo ADA disposition.
I. Philosophy
Every AI coding tool embeds a behavioral model in its system prompt — rules about when to be verbose vs. terse, when to ask vs. act, how to handle errors, what to prioritize. These behavioral models are rarely documented as such; they emerge from scattered instructions like "be concise," "always verify before executing," and "prefer simple solutions."
This skill makes the implicit explicit. By mapping these scattered instructions to Dojo's structured ADA disposition fields, we gain two things: (1) a precise vocabulary for comparing how different tools behave, and (2) a library of validated disposition presets we can import into our own agents.
The goal is not to copy other tools but to learn from the design space they have explored.
II. When to Use
- After
ingest-system-prompt has stored a system prompt as a MemorySeed
- When designing a new agent disposition and wanting to draw from proven patterns
- When comparing behavioral approaches across multiple AI tools
- When a user asks "how does Cursor/Copilot/Windsurf handle X?"
Do NOT use for analyzing Dojo's own system prompts (that is self-reflection, not external analysis) or for prompts that are not agent system prompts (marketing copy, documentation, etc.).
III. Workflow
Step 1: Retrieve the Ingested Prompt
Load the system prompt from MemoryStore using the entry ID or source tool name:
- Query:
Search(ctx, {Text: source_tool_name, EntryType: "system_prompt"}, 1)
- Validate: entry exists and has parsed sections in metadata
Step 2: Extract Behavioral Indicators
Scan the prompt text for patterns that map to ADA fields. Use this evidence table:
Pacing indicators:
| Signal | Disposition value | Confidence |
|---|
| "think step by step", "take your time", "be thorough" | deliberate | 0.8 |
| "be efficient", "avoid unnecessary output" | measured | 0.7 |
| "respond quickly", "be concise", "keep it short" | responsive | 0.8 |
| "minimize latency", "stream immediately", "no preamble" | rapid | 0.9 |
Depth indicators:
