Prompt injection mitigation: update pattern-based detection - #9198
Conversation
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 26791afd09
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 88ed696f2d
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: aa5c9ab851
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
|
Looks good! Would you please address the above codex comments before merge the PR? Thanks! |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: ccc6ff5173
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
… focus on what actually reduces FPs while still catching bad commands
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 5c01c94b65
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
* main: (102 commits) Dynamically refresh skill instructions each turn (#9217) Build non-vulkan linux variants using ubuntu 22.04 (#9211) fix(ui): show tool name in approval prompt (#9216) feat: add Atomic Chat as declarative OpenAI-compatible provider (#9210) chore: bump package.json versions from 0.19.1 to 0.20.0 (#9218) feat: support GOOSE_OAUTH_CALLBACK_PORT for stable OAuth redirect_uri (#9209) [RFC] feat(oauth): proactive token refresh to avoid re-auth on every session (#8386) fix: resolve Azure CLI on Windows by using az.cmd (#9215) fix: handle non-interactive terminal in goose configure on Windows (#9214) Better parsing of pasted html as markdown so agents understand (#9190) fix: persist accumulated cost in session DB to survive reload (#9191) fix(publish-npm): build binary from current SHA + add compat check (#9212) feat(desktop): add goose://new-session deep link to open fresh chat (#9196) Add PR previews using cloudflare pages (#9208) fix: prevent tool-use marker leakage in toolshim output (#8310) Prompt injection mitigation: update pattern-based detection (#9198) remove goose2 related skills (#9189) Switch GH pages deploy to actions/artifact workflow (#9025) fix(summon): re-apply canonical limits when delegate overrides model (#9183) Split code signing from build (#8587) ...
Summary
This PR aims to decrease false positives in pattern-based detection, especially in
rm -rfcommands. Whenever the command injection classifier is not available (i.e. not enabled in settings or WARP disabled), there is an automatic fallback to pattern-based detection. The command injection classifier has reasonably high accuracy, but the pattern-based detection is harder to get right.Changes
Estimated impact
Eliminates ~257 of ~310 weekly pattern-based flags (~83%) while all true positives remain covered by:
All other 20+ patterns unchanged (reverse shells, SUID, credential access, data exfil, etc.)
Testing
Unit tests.