Remove references to repos that aren't public

[dmccoystephenson]

Summary

Prepares create-dev-loop to be made public, by removing every reference a reader can't follow.

Step 7 (personal catalog) wrote into a hardcoded personal catalog path. It's now opt-in via $CLAUDE_SKILLS_CATALOG, with no default path — a hardcoded one would mean writing into a directory the user never asked this skill to touch. Unset, or set to something that isn't a git repo, skips the step silently, as before. Also tightened the surrounding instructions: match the existing table's column order rather than assuming this one, and skip rather than inventing a table if the catalog has none.

Validation guidance in CLAUDE.md and CONTRIBUTING.md named private reference repos as the fixtures. Reframed around any repo the contributor maintains, with a warning to prefer a low-stakes one the first time — Step 6 creates a GitHub repo, and Steps 3 and 5 write outside the project directory.

RESEARCH.md citations are anonymized rather than deleted, so the replication evidence behind each finding survives without naming an inaccessible repo. The retrofit-checklist note in CLAUDE.md no longer depends on a private CONVENTIONS.md.

Cross-linking: documented gardener as the open-source companion that dispatches generated skills across a fleet of repos, unattended and safety-gated. The split is now stated explicitly in both projects: create-dev-loop decides what a skill knows about its repo, gardener decides when and how safely one gets to run, and neither requires the other.

The consequential part of that cross-link is a new CLAUDE.md section recording that Steps 3, 5, and 6 are a load-bearing interface, not internal detail. gardener reads a generated skill from exactly the paths Step 3 and Step 5 write, and invokes /create-dev-loop itself to bootstrap a skill for a repo that lacks one. Changing where a skill is written, what it's named, or what Step 6 creates is a breaking change there and needs flagging in the PR description.

Research grounding

No finding applies — this is a docs-and-configuration change. The one edit to create-dev-loop.md (Step 7) changes where an optional catalog entry is written, not how repos are explored or how skills are structured, so no phase behavior or template placeholder is affected.

The RESEARCH.md edits are anonymization only. No finding, citation, confidence level, or Implementations entry is added, removed, or reinterpreted.

Doc sync check

• README.md's "What it does" Step list still matches create-dev-loop.md's Steps 1:1 — Step 7's description updated in both; verified by scripts/check_docs.py
• Every {{placeholder}} added or changed has a corresponding Step 4 substitution-table row — none added or changed
• RESEARCH.md updated — anonymization only, per above

Test plan

scripts/check_docs.py passes locally (also runs in CI on this PR), covering the README ↔ Steps 1:1 invariant, placeholder/substitution-table coverage, and relative links between the repo's own docs.

No behavioral validation run against a live repo, and none is needed here: the only create-dev-loop.md change is Step 7, whose entire code path is already gated behind a variable that is unset for every existing user, so the observable behavior for anyone not opting in is identical — the step skips silently, exactly as it did when the old hardcoded directory was absent. Steps 1–6, all placeholders, and the generated skill's content are untouched.

Separately verified that no reference to a non-public repo remains, by grepping both this repo and gardener against the full list of the account's private repositories. The only surviving matches are English words that happen to also be repo names.

🤖 Generated with Claude Code

---

drafted by Claude on behalf of Daniel Stephenson