What was done

Spun off from daily-repo-health-check QA item P13. Recursively scanned every repo under /Users/samaguiar/Documents/Projects/Repos/ for nested .git directories. Found 6 across 3 parent repos: 3 in sail-infrastructure/mcp-servers/ (playwright-mcp, ringcentral-mcp-2026-complete, mcp-Salesforce), 2 in sail-litify/workspaces/litify-spring-cleaning/ (claude-skills, mcp-Salesforce-), and 1 in sail-templates/Template-Refresh_project.

Earlier estimate (3 in sail-litify, 6 in sail-infrastructure) was off. Actual: 3 in sail-infrastructure, 2 in sail-litify, 1 in sail-templates.

Full report: /Users/samaguiar/Documents/Codex/repo-health/nested-git-investigation-2026-05-27.md

Why the session ended

Investigate-only run completed per task contract. Destructive cleanup deferred pending Sam's approval of the MC question below.

Key findings

• Items 3 (sail-infrastructure/mcp-servers/mcp-Salesforce) and 5 (sail-litify/.../mcp-Salesforce-) are byte-identical duplicate clones of the same fork. The sail-litify copy is a leftover from a litify-spring-cleaning staging move and can be removed wholesale.
• Item 6 (sail-templates/Template-Refresh_project) is a git init with zero commits, NOT gitignored, NOT tracked by the parent. After strip, parent will see 11 untracked files needing a Sam-call on track-vs-ignore.
• Items 1 (playwright-mcp, 514 commits) and 3 (mcp-Salesforce, 733 commits) are active forks last touched April 2026. They should become proper git submodules.
• Items 2 (ringcentral-mcp-2026-complete, 2 commits) and 4 (claude-skills, 1 commit) are vendored snapshots with thin history. Strip the inner .git, keep gitignored.
• Items 1 through 5 are already gitignored by their parents, so they are not leaking history into the parent tree. The risk is tooling confusion from walking nested .git markers (Screaming Frog crawl scripts, file-cleanup skills, indexers), not accidental commits.

Recommended next actions

1. Approve cleanup approach A, B, or C from the MC question below.
2. Once approved, run the destructive cleanup. Archive every nested directory to ~/Documents/Projects/admin/_archive/ first (tarball + sha256) for rollback.
3. After cleanup, re-run the nested-.git scan to verify count is zero.
4. Consider a follow-up sweep of ~/Documents/Projects/ outside Repos/. Original P13 scope was Repos-only; drift likely exists elsewhere too.

Handoff for next agent

Next agent does not need to re-run the scan unless more than 24 hours have passed or new repos have landed under Repos/. Just read the canonical report at /Users/samaguiar/Documents/Codex/repo-health/nested-git-investigation-2026-05-27.md and proceed from Sam's approved option.

When executing the destructive run, prefer Desktop Commander host-shell for any file mutation under ~/Documents/ per the preload guard. Do submodule conversions and the inner-.git strip back-to-back so live MCP-server checkouts (items 1 and 3) are not in a broken state between sessions.