What happens
When the review agent re-reviews a PR, it updates its sticky comment in-place with the latest findings. A separate status comment records the workflow run URL and timestamps. These two comments are not linked — a reader must manually cross-reference timestamps and head SHAs between them to determine which run produced the current findings.
What should happen
Each time the sticky comment is updated, it should be possible to trace which execution produced the current content. The provenance could point to the status comment on the PR, to the workflow run directly, or to both — whatever makes the audit trail easiest to follow.
Context
Observed on PR #481 after three consecutive review agent runs. The sticky comment body changed across runs (findings were resolved, new ones appeared) but there was no indication of which run produced the current state. The only way to trace provenance was to match the <!-- **Head SHA:** ... --> HTML comment against the status comments' timestamps.
This is a traceability gap. The sticky comment is the primary artifact readers consult for review findings, but it lacks attribution to the process that produced it. Linking the execution makes the review audit trail self-contained — no cross-referencing needed.
What happens
When the review agent re-reviews a PR, it updates its sticky comment in-place with the latest findings. A separate status comment records the workflow run URL and timestamps. These two comments are not linked — a reader must manually cross-reference timestamps and head SHAs between them to determine which run produced the current findings.
What should happen
Each time the sticky comment is updated, it should be possible to trace which execution produced the current content. The provenance could point to the status comment on the PR, to the workflow run directly, or to both — whatever makes the audit trail easiest to follow.
Context
Observed on PR #481 after three consecutive review agent runs. The sticky comment body changed across runs (findings were resolved, new ones appeared) but there was no indication of which run produced the current state. The only way to trace provenance was to match the
<!-- **Head SHA:** ... -->HTML comment against the status comments' timestamps.This is a traceability gap. The sticky comment is the primary artifact readers consult for review findings, but it lacks attribution to the process that produced it. Linking the execution makes the review audit trail self-contained — no cross-referencing needed.