dovetaillisted

# dovetail Checks whether a repository agrees with itself, and walks through what it finds. Two layers produce findings. **Exact** findings are computed in Python - links, anchors, orphans, duplicates, flag and signature drift, conventions, git-history signals. They come in two tiers. **Proven** ones follow from the structure alone, such as a link to nothing. **Heuristic** ones are likely problems that intent can explain, such as a file nothing links to. `next` prints the tier. **Judged** findings come from reviewers - contradictions, semantic staleness, spec drift, non-Python dead code. They are probabilistic. The user must always know which they are looking at. Never blur the two. **What it does not check.** Only links inside the repository are resolved: external URLs are skipped unless the user asks for `--external-links`. Links are read from `.md` and `.markdown` files, so `.mdx` pages are not checked. Say so if the user asks for either. ## Running outside Claude Code The commands below use `${CLAUDE_SKILL_DIR}`. Claude Code sets it, and `install-codex.sh` writes the real path in its place. If it is neither set nor replaced, use the directory this `SKILL.md` is in. - **No question box** (Codex, or an agent installed through skills.sh): ask in plain text instead. Print the same header, question and numbered options, one finding per message, then stop and wait for the answer. - **No model choice per subagent:** run each shard on the model you have, and say in the hea