Claude Code: Completion confirmation without output verification (last-5% reliability gap)

Behavior observed:
Claude Code confirms tasks as "done" after writing to files, without re-reading the output to verify correctness or completeness. This is especially noticeable at session end when explicitly asked to save all session state.

Concrete example:
At the end of a session, I asked Claude to document the session outcome in a Status.md file and confirm it was done. Claude confirmed. In the next session, the entry turned out to contain only the beginning-of-session state — the actual outcome (root cause found, fix deployed, architectural decision made) was missing. Claude had written something and reported completion without verifying whether the content was actually complete.

Pattern:
This happens consistently at high context (~90%) when Claude is asked to consolidate session state before /clear. The write operation is performed, but no read-back occurs. The confirmation is based on "I wrote it" not "I verified it."

Impact:
In otherwise very productive and high-quality sessions, this creates a reliability gap that forces the user to manually verify outputs that were explicitly confirmed as complete. The last 5% of accuracy is missing.

Suggestion:
Claude should re-read its own written output before confirming completion — especially for documentation tasks explicitly requested as session-closing steps.