feat(ai): add feedback skill setting#11341
Conversation
Add a default-on setting that controls whether Warp's bundled feedback skill is exposed to Oz. Gate bundled feedback skill discovery and direct bundled read_skill resolution while preserving path-based user skills and other bundled skills. Co-Authored-By: Oz <oz-agent@warp.dev>
Remove the feedback issue helper's implicit repository default and require callers to pass the target repo explicitly. Update the bundled feedback skill instructions and helper tests for explicit repo forwarding. Co-Authored-By: Oz <oz-agent@warp.dev>
|
I'm starting a first review of this pull request. You can view the conversation on Warp. I completed the review and no human review was requested for this pull request. Comment Powered by Oz |
![oz-for-oss[bot]](https://avatars.githubusercontent.com/in/3450139?s=60&v=4){kind=small}
There was a problem hiding this comment.
Overview
This PR adds a default-on Agent setting to disable Warp's built-in bundled feedback skill, wires that setting through skill discovery and direct read-skill execution, updates the Settings UI, and requires the feedback issue helper's target repository to be explicit.
Concerns
- Disabling the feedback bundled skill is currently enforced only in the
read_skillexecutor path. Other stale or explicit bundled-skill references that still resolve throughSkillManager::skill_by_referencecan continue to expose or invoke the disabled feedback skill.
Verdict
Found: 0 critical, 1 important, 0 suggestions
Request changes
Comment /oz-review on this pull request to retrigger a review (up to 3 times on the same pull request).
Powered by Oz
Co-Authored-By: Oz <oz-agent@warp.dev>
|
/oz-review |
|
I'm re-reviewing this pull request in response to a review request. You can view the conversation on Warp. I completed the review and no human review was requested for this pull request. Comment Powered by Oz |
There was a problem hiding this comment.
Overview
This PR adds a default-on Agent setting for Warp's built-in bundled feedback skill, wires the setting into bundled-skill discovery and direct skill reads, updates the Agent settings UI, and changes the feedback helper so callers must pass --repo explicitly. The description includes visual evidence for the user-facing settings change.
Concerns
- The new repo argument validation in the feedback helper should be tightened before relying on it for browser-mode issue URLs; see inline suggestion.
- No approved repository spec context was available for this review.
Verdict
Found: 0 critical, 0 important, 1 suggestions
Approve with nits
Comment /oz-review on this pull request to retrigger a review (up to 3 times on the same pull request).
Powered by Oz
| if not normalized_repo or "/" not in normalized_repo: | ||
| raise SystemExit("Issue repo must be provided in owner/name format.") | ||
| owner, name = normalized_repo.split("/", 1) | ||
| if not owner or not name or "/" in name: |
There was a problem hiding this comment.
💡 [SUGGESTION] [SECURITY] repo is later interpolated into a GitHub URL, but this validation still accepts URL/path control characters such as ?, #, spaces, or control characters inside owner/name; validate both components against GitHub's allowed repository-name character set before returning it.
There was a problem hiding this comment.
Don't think this is a huge deal and probably OK to skip, but wouldn't hurt. Will leave up to you.
There was a problem hiding this comment.
I agree -- worst case scenario the agent has to call the script again with the correct format.
Co-Authored-By: Oz <oz-agent@warp.dev>
| parser.add_argument( | ||
| "--repo", | ||
| required=True, | ||
| help="GitHub repository to file the issue in, formatted as owner/name.", | ||
| ) |
vkodithala
left a comment
vkodithala
left a comment
There was a problem hiding this comment.
LGTM! Thanks for the fix 🚀
captainsafia
deleted the
safia/app-4549-add-setting-to-disable-the-feedback-skill
branch
2 participants
Description
Adds a default-on Agent setting to disable Warp's built-in bundled
feedbackskill without affecting user-created skills or other bundled skills.This PR also hardens the bundled feedback issue helper in a separate commit so callers must pass the target GitHub repo explicitly with
--repo; the normal Warp feedback target remainswarpdotdev/warpin the skill instructions.Commits are intentionally split:
feat(ai): add feedback skill settingfix(ai): require repo for feedback issuesLinked Issue
ready-to-specorready-to-implement.Testing
./script/runpython3 resources/bundled/skills/feedback/scripts/test_file_feedback_issue.pypython3 resources/bundled/skills/feedback/scripts/file_feedback_issue.py --help--repoargparse check forfile_feedback_issue.pycargo fmtcargo test -p warp bundled_feedback_skillcargo test -p warp disabling_feedback_bundled_skillcargo test -p warp read_skill_executorcargo test -p warp get_skills_for_working_directorycargo test -p warp best_supported_providercargo test -p warp cloud_environment_skills_always_includedcargo test -p warp test_read_bundled_skillscargo test -p warp ai_page_testscargo clippy --workspace --all-targets --all-features --tests -- -D warningsgit --no-pager diff --checkScreenshots / Videos
Demo Loom: https://www.loom.com/share/f8d406d0237d46ac8b5ffbe674583fb4
Agent Mode
Agent artifacts
CHANGELOG-OZ: Added a setting to disable Warp's built-in feedback skill.
Co-Authored-By: Oz oz-agent@warp.dev