fix(source-control): name --keep-data in setup uninstall guidance (0.55.6)

[kyle-sexton]

Closes #3131

Summary

/source-control:setup's babysit-config section warned operators not to uninstall the plugin in order to reconfigure it, citing the pluginConfigs entry an uninstall drops, and stopped there. --keep-data appeared nowhere in the plugin. An operator who uninstalls for a legitimate reason (troubleshooting, changing scopes, reinstalling a version) had no warning that uninstalling from the last remaining scope deletes ${CLAUDE_PLUGIN_DATA} by default, where this plugin stores babysit-prs state and, at the last resolution rung, worktrees.

Fix

One paragraph added to skills/setup/SKILL.md's babysit-config section, after the reconfigure bullets. It names --keep-data, states what the directory holds, and states which resolution rung each worktree root must fall through before it lands there.

What the directory actually holds, verified in-tree:

Path	Contents	Relocatable?
${CLAUDE_PLUGIN_DATA}/state/babysit-prs	queue state, worker leases, feedback ledger	No userConfig key relocates it
${CLAUDE_PLUGIN_DATA}/worktrees	babysit worktrees; /source-control:worktree trees	Only at the last resolution rung

babysit_worktree_root falls back there whenever it is unset. /source-control:worktree create reaches that directory only when neither the target repository's melodic.worktreeroot git config nor worktree_root resolves (worktree-create.sh rungs 2 and 3). Babysit's own worktrees are ephemeral scratch that rebuild from GitHub; the state directory and a /source-control:worktree tree holding uncommitted work are not.

The existing guidance against uninstalling purely to reconfigure is unchanged. Plugin version 0.55.5 → 0.55.6. Follows docs/conventions/plugin-data-report-keying/README.md Rule 4.

Verification

• Official docs re-fetched at rung 1 on 2026-08-23 (https://code.claude.com/docs/en/plugins-reference.md, 108305 bytes): "By default, uninstalling from the last remaining scope also deletes the plugin's ${CLAUDE_PLUGIN_DATA} directory. Use --keep-data to preserve it." The plugin uninstall flag table carries --keep-data spelled that way.
• rg -n 'keep-data' plugins/source-control/ now hits the new paragraph and the changelog entry (zero hits on origin/main).
• Existing "Do not uninstall to reconfigure" warning remains intact at setup/SKILL.md:254.
• claude plugin validate plugins/source-control: Validation passed.
• scripts/check-changed-skills.sh origin/main (setup): PASS — 0 errors, 1 pre-existing soft-target warning.
• scripts/check-changelog-parity.sh --check / --check-bump origin/main / --check-order / --check-preserved origin/main: PASS.
• scripts/sync-plugin-options-docs.py --check: up to date.
• scripts/check-skill-count-claims.sh --check: PASS.
• scripts/check-skill-leaf-names.sh --check: PASS.
• scripts/check-cross-plugin-source-drift.sh --check: PASS.
• markdownlint-cli2 on both changed markdown files: 0 issues.

Related

• Closes source-control/setup: uninstall guidance omits --keep-data (destroys babysit worktrees and lease state) #3131 — F3 from the /plugin-quality:audit packet. The other findings in that packet are out of scope and tracked separately.
• Refs Cross-plugin: hook-utils skip latch silences 9 formatters, overlay-ignore guard unreachable, plugin-quality seal premise false #3128 — F1 from the same run, shared with work-items.
• Refs fix(source-control): name --keep-data in setup's uninstall guidance (0.55.6) #3211 — earlier close-linked PR for this issue; it conflicts with origin/main after source-control 0.55.5 landed (source-control: worktrees created outside worktree-create.sh carry no lock, so concurrent sessions reach into each other's trees with no claim to check #2882 / fix(source-control): claim unlocked worktrees created outside the helper #3104). This PR is the rebase onto current main at 0.55.6.
• The generated plugin-README uninstall warning is left alone: it lives in the BEGIN GENERATED block from scripts/sync-plugin-options-docs.py and a hand-edit would be clobbered. Adding --keep-data there is a marketplace-wide edit, not a source-control one.

[claude]

Claude finished @kyle-sexton's task in 25s —— View job

---

Security review complete ✅

• Gather context (PR diff, changed files)
• Apply skip gate
• Review changed files for security-relevant surface
• Post findings or clean-review summary

Result: No security issues found.

This PR only touches documentation and metadata:

• plugins/source-control/.claude-plugin/plugin.json — version bump 0.55.5 → 0.55.6
• plugins/source-control/CHANGELOG.md — new changelog entry
• plugins/source-control/skills/setup/SKILL.md — one added paragraph in the babysit-config section documenting claude plugin uninstall --keep-data

There's no code execution, no GitHub Actions workflow changes, no user input handling, no credential/token handling, and no injection surface (command, SQL, path, template) introduced or modified. The skip gate in /review:security-review applies: this change has no security-relevant surface.

[github-actions]

Last security-reviewed head: 90a551399f3bc62537b8c5d59d1a9fe2561c3c32. On the next push, the relevance gate compares only the commits since this SHA; delete this comment to force a full re-review.

[kyle-sexton]

Closing this as superseded by #3211. Both pull requests close-link issue #3131 and carry identical changes (the same three-file diff: the version bump to 0.55.6, the CHANGELOG entry, and the --keep-data paragraph in the setup skill). Per this org's duplicate-resolution convention the lowest-numbered PR is canonical, so #3211 — opened earlier at 19:30 UTC — carries the fix forward. Nothing here is lost: #3211 contains the same content in full. Duplicates like this are a normal outcome of parallel lanes picking up the same issue; thanks for the fix.

[claude]

Claude finished @kyle-sexton's task in 23s —— View job

---

Code review — skipped

• Gather context
• Determine review scope
• Apply /review:code-review skip gate

Result: No review performed. This PR is closed (not open), which trips the skip gate in /review:code-review: "PR is closed or not open → stop early, say so in the summary, post nothing else." No inline comments were posted.

If this PR is reopened, re-trigger @claude and I'll run the full review lane.

[github-actions]

Claude has reviewed this PR 1 time. The lane skips further automatic reviews after 5; deleting this comment resets the count.