recipe: Maverick — behavioral adaptation for Goose#8058
recipe: Maverick — behavioral adaptation for Goose#8058smartrickpicks wants to merge 1 commit intoblock:mainfrom
Conversation
Adds an MCP-based recipe that adapts Goose's communication style to match each developer's behavioral profile. 10-question assessment detects working style across four drives (Dominance, Extraversion, Patience, Formality), then adjusts pacing, tone, autonomy, and decision style accordingly. Features: - Profile detection via behavioral assessment - Adaptive communication (response length, tone, checkpoint frequency) - Shareable Developer Card visualization - Local-only storage (~/.maverick/), zero telemetry - Published on PyPI: uvx maverick-mcp serve Repository: https://github.com/get-airlock/maverick-mcp
|
🔍 Recipe Security Scan Results 📊 Scan Summary:
📋 Individual Recipe Results: 🔗 View detailed scan results in the workflow artifacts. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: bfcfd42503
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| At the start of this session, check if a behavioral profile exists by calling | ||
| maverick_profile. If no profile exists, offer to run a quick 2-minute assessment | ||
| by calling maverick_assess (no arguments to get the questions, then call again |
There was a problem hiding this comment.
For stdio extensions Goose exposes tools as <extension>__<tool> (see fetch_all_tools in crates/goose/src/agents/extension_manager.rs, which prefixes every non-platform extension). This recipe instead tells the model to call maverick_profile, maverick_assess, maverick_adapt, and maverick_feedback, so the very first session bootstrap will hit “Tool not found” instead of invoking Maverick whenever the recipe is used.
Useful? React with 👍 / 👎.
| - Decision-making style (decisive vs consultative vs evidence-based) | ||
| - Checkpoint frequency (how often to check in) | ||
|
|
||
| After assessment, offer to generate a Developer Card by calling maverick_card. |
There was a problem hiding this comment.
maverick_card tool call
I checked the published maverick-mcp 0.1.0 package description on PyPI: its MCP tool list contains only maverick_assess, maverick_profile, maverick_adapt, and maverick_feedback. maverick_card appears to be a CLI feature, not an MCP tool, so asking Goose to call it will always fail with a missing-tool error for anyone who tries to generate the Developer Card from this recipe.
Useful? React with 👍 / 👎.
|
Thanks for the recipe — the concept is solid and the overall structure follows existing recipes well. A few things to fix before we can merge: 1. MCP tool names need the goose exposes stdio extension tools as
As written, every tool call in the prompt would fail with "Tool not found". 2. The prompt instructs goose to 3. Brand casing: Per our style guide, the product name should always be lowercase Once those are addressed we can get this merged! |
DOsinga
added
the
needs_human
2 participants
Summary
Adds a recipe for Maverick — an MCP server that adapts Goose's communication style to match each developer's behavioral profile.
~/.maverick/, zero telemetry, no accountsPublished on PyPI:
uvx maverick-mcp serveHow it works
The recipe uses
uvxto run the server — no manual install needed.Test plan
uvx maverick-mcp serveand verified MCP tools respond~/.maverick/card.htmlsearch-datahub.yaml)Every Goose needs its Maverick.