Repair ETag Precondition Races

The problem

Race conditions occur when HTTP If-Match checks are evaluated outside of the database transaction, leading to lost updates or stale data corruption. Weak ETags are often incorrectly accepted for unsafe methods, or 304/412 status codes are reversed, breaking client-side caching logic.

What it does

• Audits repository code to find non-atomic precondition checks in read-modify-write cycles.
• Refactors ETag generation to ensure validators accurately identify specific representation versions.
• Standardizes strong versus weak comparison logic and wildcard semantics across create and update paths.
• Implements deterministic local fixtures and barriers to verify concurrency fixes without production access.
• Injects redacted observability to track rejections and terminal state transitions.

Why this beats prompting it yourself

General LLMs often suggest simple retries or lock-waits that don't address the underlying architectural flaw of non-atomic validation. This skill enforces a strict resource register and failure timeline, ensuring the fix is bound to the transaction lifecycle rather than just adding "optimistic" fluff.

Use cases

• Preventing two concurrent writes from both passing a stale If-Match precondition.
• Correcting repository logic where 304 Not Modified is sent for changed resources.
• Hardening API endpoints that allow weak ETags for PUT or PATCH operations.
• Fixing inconsistent wildcard handling between resource creation and updates.

Known limitations

Requires access to local repository files to inspect logic. It cannot fix issues residing in external load balancers, CDNs, or third-party API gateways that strip headers.