[BUG] Factory reviews get stuck on GitHub App author name formats

Finding: In an external report (not reproduced by KFA), Factory's factory-review and factory-rereview skills withhold review publication and the recorded verdict for GitHub App-authored PRs, because they compare two formats of the same author name. Source: https://github.com/mastra-ai/mastra/issues/26496. The reporter says the behavior is present in installed @mastra/factory 0.19.1, published 0.20.0, and upstream main at 7a969517; the triage comment says it reproduced on main 25b8b607. Reported mechanism: the gh CLI command `gh pr view --json author` returns `app/mastra-platform` for PR 26452, while the REST pulls response gives `mastra-platform[bot]` in user.login. Factory records the REST login at intake. The skill text then compares it with the CLI login and treats the mismatch as reviewing the wrong PR (issue body). The reporter's read-only script prints both values and then fails its equality assertion; the reporter notes this reproduces the prescribed comparison, not a full model-driven review, although a real local review also withheld publication after its technical checks. The automated triage comment (model synthesis, not independent evidence) traces the intake author to the REST payload in Factory GitHub integration code and locates the comparison at factory-review SKILL.md line 141 and factory-rereview line 148. It rates the severity as medium, with human-authored PRs unaffected, and says App/bot-authored PRs fail the cross-check even when repository, URL and head SHA match (https://github.com/mastra-ai/mastra/issues/26496#issuecomment-6100511665). It also suggests that the separate author-identity misconfiguration step, which compares gh api user to the PR author, may have the same format mismatch and could hide a self-review. Proposed direction, not shipped: fetch the author for the cross-check from the same REST endpoint used at intake, keep failed lookup, missing author or genuinely different author as a hard stop, preserve the repository, PR-target, provenance and head-commit checks, and do not rewrite app/<slug> into <slug>[bot]. The reporter says a local fix with skill-contract tests exists (12 focused tests passing) but has not opened the PR pending approval. Status: the issue is OPEN, no merged fix or release containing a fix is identified in the supplied sources, and the only stated workaround is manual review. Limitation: this is an external report with no KFA reproduction; a full model-driven review has not been rerun with the fix.