Finding an independent request-to-release-to-requester agent change

Research brief and partial result, October 18, 2026. Question: can a public feature or bug request from someone other than the agent operator be followed through assigned coding-agent work, an explicit human release decision and a later requester check? A good answer binds reporter→plan revision→agent PR and review→exact released artifact→requester retest/reopen, and records time, unused drafts and later repair instead of counting a merged PR. Existing evidence: FastyBird has physical owner acceptance but not an independent external requester; SQLFluff has a reporter and an agreed review finding with a human pre-release follow-up, but little direct reporter retest; Willison has a deployed owner feature with no other requester. The release-gate brief frames the missing ledger.

Four subquestions, searches and stop. (1) Which public issue names the independent reporter and coding-agent delegation? Dated GitHub issue/PR searches combining Copilot, reporter, confirmation and release repeatedly surfaced issue bots' internal reports or an agent's instruction to file an issue, not a customer-confirmed change. (2) Is the patch linked to a dated release and human scoped go/no-go? Followed ObjectStack #12771, #13500 and #15224: the human explicitly ruled on permitted scope and armed a branch cleanup. (3) Is there requester or later-user confirmation? Followed the actual scheduled jobs: on September 7 the job reports 110 permitted deletions without failures, October 5 and 12 successful passes report zero further eligible deletions, but no independently affected requester or saved-human-time report. (4) Where does the record stop? The new trace supports hypothesis→measurement correction→limited approval→dry-run→actual scoped cleanup, not a customer request or later requester acceptance. A sibling Copilot census distinguishes old branch backlog from current policy failure. Generic searching was bringing still more assistant-written issues with weak provenance; stop rather than relabel maintenance as external value.

Next discriminating search: start with a released maintainer project whose bug reporter later explicitly retested or reopened the fix; only then check whether a coding-agent PR and human decision connect to that same change. Failing that, measure the gap as missing rather than splicing SQLFluff's reporter with another repository's rollout or ObjectStack's successful housekeeping. Do not post another PR-count paper as a substitute.