Priority and assignment during capture review

Intent

Reviewers can choose a priority and one responsible assignee before sending screenshots or video, without opening the resulting thread. The selected member opens a dropdown containing its own search field and at most ten members per page. More/Previous are compact paging controls; Unassigned is a list choice. Keyboard selection and outside-click/Escape dismissal remain available. Unassigned and Normal are the defaults. Group/team administration and assigning the same whole task to several developers are outside this change.

Boundaries

• Reuse durable assignments and work plans, with server authorization and atomic thread creation.
• Eligible assignees are active writable members of the selected project or organization owners. Return a bounded name/id list without profiles, email or policy details.
• Preserve approved draft metadata and exact retry identity. Clear an assignee when changing projects.
• Existing extension credentials retain their scopes. Newly approved pairing includes member lookup, planning and assignment; old connections receive reconnect guidance.
• Keep GitHub decisions undecided. Assignment does not start another member's agent or create an external issue.
• Native Chromium recording exposed an existing upload rejection for WebM BlockGroup/Block frames. Support this documented WebM representation, while retaining header, video-track, media and malformed-boundary checks (checked 2026-10-01).
• Server and extension versions are separate; a source package is not Chrome Store publication.

Verification

• Atomic creation, idempotent retries, denied scopes, stale/ineligible assignees and rollback.
• Hundred-member search/paging, minimal fields and project isolation.
• Actual screenshot and video editor integration, keyboard, responsive UI, project changes and frozen retries.
• Appropriate focused checks, repository gate and package inspection.

Progress

Source implementation and package acceptance are complete. Node 24 repository checks passed with 507 tests and 34 optional skips; the packaged editors verified a hundred-member project, internal search, paging, keyboard and outside-click/Escape controls, persisted screenshot drafts and exact screenshot/video retries. The native-video scenario uploads real MediaRecorder WebM through the isolated server and confirms one thread/assignment after a lost create acknowledgement. Full extension regression and required CI remain release gates. Hosted acceptance and Chrome Store publication require separate rollout evidence.