Methodology

Every record carries one of four confidence levels, derived from its evidence:

confirmed-official

Confirmed by official Microsoft sources — documentation, error-code references, or an explicit Microsoft statement.

corroborated

Confirmed by at least two independent non-official sources (different authors, not derived from each other).

single-report

Based on a single credible report, such as one community thread or one practitioner write-up. Treat with care.

practitioner-opinion

An experienced practitioner's assessment without a documented failing case. The weakest level; never used for hard claims about limits.

Every record shows "Last verified" — the date of its most recent verification event. There are three kinds:

doc-check

The cited official documentation was re-read and still supports the record's claims.

community-check

Recent community threads were reviewed for new reports, fixes, or changes in behaviour.

own-repro

The problem was reproduced (or no longer reproduced) in our own test tenant, with date and environment context. "Reproduced by us" is the strongest trust signal in the register and is rendered prominently on the record.

Each record carries one status: open (the problem exists), mitigated (a reliable workaround exists), fixed (Microsoft resolved it — and we verified that ourselves; a release note alone is never enough), or regressed.

Regressed means a problem that was fixed has come back. It stays visible as its own status with its full history — a record is never silently reset to open. Fixed records are spot-checked on an ongoing basis, because every platform update is a potential regression trigger.

A status can be flagged as partial when it only applies to part of the affected surface, e.g. fixed for one method but still open for another. The badge then carries a one-sentence note.

Errors in the register itself are corrected in the open: every substantive change lands in the record's visible changelog, dated. There is no silent editing. Numbers and limits in the text always name their source and an as-of date.

This register is not affiliated with, endorsed by, or paid by Microsoft. There are no affiliate links and no advertising in records. Criticism is evidence-based and constructive; the register exists to get you unstuck, not to bash a product. It is published by PowerLeap Advisory GmbH (Switzerland), which offers consulting around Copilot Studio — the register itself is and stays fully open.

Live numbers, recomputed on every build, on how well the register keeps its own verification promise:

Records in the register

40

Published friction records — drafts and unreviewed pipeline output never count.

Verified within 90 days

100%

Share of records whose most recent verification event is at most 90 days old.

Median verification age

12 days

Median time since each record was last checked against docs, community, or our own repro.

Status distribution

4 open · 2 regressed · 34 mitigated

How the register splits across open, regressed, mitigated, and fixed right now.