Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 10 additions & 0 deletions docs/security.md
Original file line number Diff line number Diff line change
Expand Up @@ -13,6 +13,16 @@
- **No Cross-Repository Access**: Each action invocation is limited to the repository where it was triggered
- **Limited Scope**: The token cannot access other repositories or perform actions beyond the configured permissions

## Pull Request Creation

In its default configuration, **Claude does not create pull requests automatically** when responding to `@claude` mentions. Instead:

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Would it be helpful to list what config controls this behavior?


- Claude commits code changes to a new branch
- Claude provides a **link to the GitHub PR creation page** in its response
- **The user must click the link and create the PR themselves**, ensuring human oversight before any code is proposed for merging
Comment on lines +18 to +22
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The phrasing "ensuring human oversight before any code is proposed for merging" could be clearer about what this oversight entails. Currently, users click a link to a pre-filled PR creation form (with title, body, and branch already configured). The oversight is confirmatory rather than preventative.

Consider rephrasing to be more precise:

The user must click the link to confirm PR creation, allowing them to review the pre-filled PR details before submission.

This more accurately describes the security boundary while still conveying the key point that automatic PR creation doesn't happen.


This design ensures that users retain full control over what pull requests are created and can review the changes before initiating the PR workflow.

## ⚠️ Prompt Injection Risks

**Beware of potential hidden markdown when tagging Claude on untrusted content.** External contributors may include hidden instructions through HTML comments, invisible characters, hidden attributes, or other techniques. The action sanitizes content by stripping HTML comments, invisible characters, markdown image alt text, hidden HTML attributes, and HTML entities, but new bypass techniques may emerge. We recommend reviewing the raw content of all input coming from external contributors before allowing Claude to process it.
Expand Down
Loading