babysit-prs: resolve babysit_self_logins discovery contract (code drops discovery-widening; plugin.json userConfig still claims it)

[kyle-sexton]

Surfaced by Codex on PR #882 (fix/511-babysit-self-identity-decouple): discussion r3624440917. Tracking here because #882 closes #511, so the fork below needs a live home.

The fork (author decision)

PR #882 decouples the self-identity suppression set from the --author discovery filter. Post-#882, babysit_self_logins is routed only to --extra-self (self-comment suppression + same-login classification + readiness-gate rows + merge-gate self-exemption) and is no longer joined onto --author. Consequence: safe/worker queue cycles no longer discover PRs authored by those configured identities.

But the config contract still advertises discovery. plugins/source-control/.claude-plugin/plugin.json babysit_self_logins userConfig description lists "the self set babysit-prs uses for discovery scope, readiness-gate classification rows, and the merge-gate self-exemption" — with discovery scope first. Code and documented contract now disagree.

Two mutually exclusive resolutions:

• A) Discovery-widening was intended -> restore those identities in --author for non-autopilot discovery (code change). Downside: partially re-couples what babysit-prs: decouple self-identity from --author so configured self-login extras survive autopilot widening #511 set out to decouple.
• B) Discovery-widening was an incidental side effect of the old author-derived self set -> drop the "discovery scope" claim from the plugin.json description (doc-only), formally making babysit_self_logins suppression/classification-only.

Why it needs an author call, not a blitz fix

#511's title frames the goal as the suppression set surviving autopilot widening, not stopping discovery of those identities' PRs. So the discovery-drop reads as an implementation side effect rather than a stated goal — but plugin.json listing discovery scope first is evidence the widening may have been intentional. Picking A vs. B embeds a design decision the author owns; a blitz worker should not settle it.

Pointers

• plugins/source-control/.claude-plugin/plugin.json — babysit_self_logins userConfig description (the stale "discovery scope" claim)
• plugins/source-control/skills/babysit-prs/SKILL.md — step-4 invocation (~L388, already updated: "self-suppression no longer rides on it") and userConfig table (~L292)
• plugins/source-control/skills/babysit-prs/scripts/pr_queue_snapshot.py — resolve_self_logins / build_config / build_snapshot

This was generated by AI (tower blitz, session 6).

[kyle-sexton]

Author ruling (from #511's author): Option B — discovery-drop is intended.

babysit_self_logins is a suppression/classification/merge-exemption set, not a discovery filter. The pre-#511 --author @me,<self-logins> widening was an incidental side effect of the old author-derived self set, not a stated goal — so it is intentionally dropped, not restored.

Basis (#511's own body is decisive):

• It frames self_logins as "the posting identities" and, under Root asymmetry, states: "self_logins (whose comments to suppress = the posting identities) is a distinct concern from --author (which authors' PRs to discover). The snapshot conflates them."
• Its Suggested fix direction explicitly says to "drop the --author @me,<self-logins> extension guidance."

So the "distinct concern" the fix set out to enforce is exactly that self-identity ≠ discovery. Which authors' PRs the queue discovers stays --author's job; if an operator wants a bot's PRs discovered, that's an explicit --author concern, independent of this set. The plugin.json description listing "discovery scope" first was describing the pre-#511 conflated behavior — stale, not evidence of intent.

Resolution (doc-only, no code change — the code already behaves this way post-#511): corrected the babysit_self_logins userConfig description on the #882 branch (commit 23de05c2) to state it is suppression/classification/merge-exemption-only and explicitly "not a discovery filter." A parallel commit on #882 had already independently dropped "discovery scope" from the description (convergent decision); this makes the full contract statement authoritative. CHANGELOG 0.16.0 (#511) entry records the ruling.

Resolves when #882 merges. Leaving open for the merge lane to close on merge.

[kyle-sexton]

This was generated by AI during triage.

Triage → needs-human + wayfind: design. Genuine author-owned config-contract decision, correctly not a blitz fix. Post-#882 (closes #511), babysit_self_logins is routed only to --extra-self (suppression/classification/merge-gate) and no longer joined onto --author, so safe/worker cycles stop DISCOVERING PRs authored by those identities — but the plugin.json userConfig description still lists "discovery scope" first. Code and documented contract disagree.

Two mutually-exclusive resolutions, with balanced evidence — which is exactly why this is human-gated (product intent on what the option is FOR):

• A) Discovery-widening intended → restore those identities in --author for non-autopilot discovery (code change; partially re-couples what babysit-prs: decouple self-identity from --author so configured self-login extras survive autopilot widening #511 decoupled). Evidence: plugin.json lists discovery scope first.
• B) Discovery-widening incidental → drop the "discovery scope" claim from the description (doc-only), formally making babysit_self_logins suppression/classification-only. Evidence: babysit-prs: decouple self-identity from --author so configured self-login extras survive autopilot widening #511's stated goal was the suppression set surviving autopilot widening, NOT discovery — so the drop reads as a side effect.

Not decision-defaulted: the reporter (full context) deliberately declined to recommend, and the evidence genuinely cuts both ways — the maintainer owns what the config option means. Sibling: #881 (same #511/#882 self-identity review — a separate fork: fail-loud vs --extra-self parity; route each independently, no shared single decision). Priority medium (source-control priority-set; reporter severity low — a contract/doc mismatch, not a live defect). Type Bug (code ≠ documented contract until reconciled either way).

(Edited to restore a term dropped by a shell-substitution slip in the original post — no routing change.)

[kyle-sexton]

This was generated by AI (decide lane — application-ready / reconciliation).

DECIDED and APPLIED — this is closable, not an open fork

Author ruling. Comment 5038158047 (19:24Z) is an explicit ruling from #511's author: Option B — discovery-drop is intended. babysit_self_logins is a suppression/classification/merge-exemption set, not a discovery filter; the pre-#511 --author @me,<self-logins> widening was an incidental side effect, intentionally dropped, not restored. Resolution is doc-only — the plugin.json babysit_self_logins description was corrected on the #882 branch (commit 23de05c2).

Now applied. PR #882 MERGED 2026-07-21T19:29:15Z — the corrected description is on main. The code≠documented-contract mismatch this issue tracked is resolved.

Stale re-open superseded. The 19:30:01Z triage comment re-presented A-vs-B as an open needs-human fork — but it posted ~6 min after the author ruling AND after #882 merged. It is superseded; there is no open decision here.

Recommended routing (triage / merge lane): clear needs-human + wayfind: design and close #897 as resolved by #882 (the author explicitly left it open only "for the merge lane to close on merge," and #882 is now merged).

Cross-links: #882 (merged carrier, 23de05c2) · #511 (origin) · #881 (sibling #511/#882 fork — fail-loud vs --extra-self parity; still separately open, route independently).