- Home
- Skills
- Agents & Orchestration
- codewalk
More screenshots
Works with the AI Agents you already use
codewalk
Generate interactive, visual HTML walkthroughs of Python codebases by recording real program executions.
$1.43
codewalk
Example session with this skill installed
Create a codewalk walkthrough for the data cleaning pipeline in main.py. Show the dataframe before and after the drop_na function.
- Read your context and instructions
- Compiled the codewalk
I have analyzed the data cleaning pipeline in main.py. I generated a walkthrough.py script that hooks into the process_data function, captures the state of the dataframe, and maps it to a three-stage schema: Load, Clean, and Export.
walkthrough.py
from codewalk import Walkthrough
walkthrough = Walkthrough(
title="Data Cleaning Pipeline Walkthrough",
entry="main.py --sample-size 100",
schema=["Load", "Clean", "Export"],
description="This walk shows how we handle missing values and outliers in the initial ingest."
)
walkthrough.step(
id="pre-clean",
title="Raw Data Ingest",
level=1,
at="main.py::process_data",
highlight="load",
show={"raw_df": "df.head(10)", "initial_count": "len(df)"},
text="The pipeline starts by loading the raw CSV. Note the {{initial_count}} rows in the 'raw_df' tab."
)
walkthrough.step(
id="post-clean",
title="Dropped Nulls",
level=1,
after="main.py:45", # Line after df.dropna()
highlight="clean",
show={"clean_df": "df.head(10)", "null_count": "df.isnull().sum().sum()"},
text="After running dropna, we have {{null_count}} missing values remaining. The cleaned dataset is now ready for the model."
)
Next steps
- Install dependencies:
pip install 'codewalk[screenshot]' && playwright install chromium. - Generate the report:
codewalk record walkthrough.py -o out. - Preview the interactive page:
codewalk serve out.
Connects securely to your tools. The creator never sees your data.
About this skill
Onboarding a new developer to a complex Python pipeline often involves hours of manual screen sharing and stale documentation. This skill automates the creation of interactive, visual walkthroughs by recording a live execution of your code under a debugger.
What it does
- Live Debugging Capture. Records a Python run once and pauses at defined breakpoints to capture state.
- Visual DSL. Uses a Python-based configuration to define steps, schema highlights, and data visualizations.
- Data Serialization. Automatically renders DataFrames as tables, tensors as stats previews, and Matplotlib figures as images.
- Multi-level Navigation. Organizes steps into hierarchical levels from high-level overview to line-by-line implementation details.
- Self-contained HTML. Generates a portable index.html file that functions as a rich, interactive playback of the code run.
How it works
- Scaffold. Run
codewalk initto generate awalkthrough.pytemplate for your target codebase. - Configure. Define breakpoints, choose expressions to
show, and draw the project schema using Mermaid or lists. - Record. Execute
codewalk recordto trigger the debugger, capture runtime objects, and generate the static site. - Validate. Use
codewalk checkto verify breakpoints andcodewalk screenshotto preview how the UI looks.
Frameworks & tools
Python, Playwright, Matplotlib, Pandas, NumPy, Torch, and Mermaid.js. Works with any standard Python environment using pip or uv.
Why this beats prompting it yourself
Writing documentation manually requires keeping code and screenshots in sync. This tool ensures that every value, plot, and code snippet shown to the reader is pulled directly from a real execution, eliminating stale documentation debt.
Use cases
- Engineer Onboarding. Create a self-paced guide for new hires to walk through the core logic of a complex service.
- Pipeline Auditing. Document data transformations in a ML pipeline by showing state before and after cleaning steps.
- Bug Reporting. Capture a failing run with full visual context of live objects and stack traces.
Known limitations
The target execution must finish in under 600 seconds by default. Breaking points on blank lines, comments, or non-executable decorators will fail to fire.
Execution must finish within a timeout (default 600s). Breakpoints on blank lines or comments will not fire. Requires Playwright for screenshots.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 11 days ago
Requires
Permissions
Allowed Hosts