2.5k
GitHub Stars
4
Bundled Files
2 months ago
Catalog Refreshed
3 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill openclaw/skills --skill chatgpt-skill- _meta.json292 B
- README.md1.8 KB
- requirements.txt111 B
- SKILL.md5.0 KB
Overview
This skill automates ChatGPT Web interactions by driving a local browser session with a persistent login state. It supports one-shot prompts, reusable multi-turn sessions, and stores conversation and session metadata for reliable reuse and troubleshooting.
How this skill works
The skill controls a local browser profile to interact with the ChatGPT web UI, preserving cookies and storage under the skill directory. It provides command wrappers to manage authentication, create and reuse sessions, send questions to specific conversation IDs, and capture debug screenshots and structured error payloads for automation resilience.
When to use it
- When you need ChatGPT-style dialogue but must use the web UI instead of the API.
- When you want persistent login and browser state across runs for multi-turn conversations.
- When automating browser-based tasks that require visible debugging (screenshots, show-browser).
- When you need to continue a known conversation ID or reuse a session across processes.
- When troubleshooting UI selector changes or intermediate login flows that require manual intervention.
Best practices
- Always run commands through the provided python wrapper to ensure the virtualenv and dependencies are correct.
- Validate authentication before deeper automation to detect redirects, 2FA, or account chooser flows.
- Use persistent sessions for multi-turn workflows and one-shot calls for ephemeral queries.
- Enable --show-browser during UI changes to inspect the live page and capture failing flows.
- Keep proxy and network environment variables explicit on Linux to avoid subtle network failures.
Example use cases
- Start a long-lived browser session and ask a sequence of follow-up questions while preserving context.
- Automate one-off queries via ask_chatgpt.py for tasks that require the web-only interface.
- Reopen and continue an archived conversation using a saved conversation ID.
- Run visible debug runs with screenshots when selectors break after a UI update.
- Create session daemons for concurrent tools that need to reuse an authenticated tab.
FAQ
All auth metadata and browser state are kept inside the skill directory under data/auth_info.json and data/browser_state/.
What happens if ChatGPT's page structure changes?
The skill returns explicit machine-readable errors and saves screenshots; use --show-browser and inspect data/screenshots/ while updating selectors.