The 2026-08-30 v1/v2 RCA found two remaining shared Flow-discovery assumptions in the matched regression cohort:
- project-scoped checks refuse two byte-identical substantive solution copies, even though there is no semantic ambiguity (observed on
skill-flow-e2e-escalation-slack-alert)
- file-scoped checks prefer a one-node solution scaffold over a substantive root SDK emit (observed on
skill-flow-webhook-waitfor-parallel)
The other observed failures remain semantic, runtime, or authoring failures: for example, pagination telemetry is an intentional requirement of the pagination-specific task and its delivered Slack binding was also incorrect.
Acceptance criteria:
- deterministically collapse byte-identical Flow project candidates for project-scoped checks
- prefer a substantive root SDK emit over abandoned project scaffolds for file-scoped checks only
- retain the current conservative refusal for genuinely distinct substantive projects and unreadable candidates
- add negative-control tests and keep all Maestro Flow checker tests green
Source evidence: /home/tmatup/root/workspaces/runs/rca-adhoc-2026-08-30_04-43-27-2026-08-30_17-49-32/official-evidence/adhoc-2026-08-30_04-43-27/v2.
The 2026-08-30 v1/v2 RCA found two remaining shared Flow-discovery assumptions in the matched regression cohort:
skill-flow-e2e-escalation-slack-alert)skill-flow-webhook-waitfor-parallel)The other observed failures remain semantic, runtime, or authoring failures: for example, pagination telemetry is an intentional requirement of the pagination-specific task and its delivered Slack binding was also incorrect.
Acceptance criteria:
Source evidence:
/home/tmatup/root/workspaces/runs/rca-adhoc-2026-08-30_04-43-27-2026-08-30_17-49-32/official-evidence/adhoc-2026-08-30_04-43-27/v2.