Filetree

Filetree Skill — Shared Rules

Cross-cutting rules used by /filetree:init and /filetree:update. The commands themselves contain step-by-step flows; this file holds rules that apply across modes so they're maintained in one place.

/filetree:lint is read-only script invocation and does not need these rules.

---

Summary style

One line, max 25 words, describes what the file is FOR (its role / purpose). Not what it implements internally.

• Good: "JWT auth middleware; parses token from request header and injects user_id into context"
• Bad: "Defines AuthMiddleware class with init and call methods"
• Bad: "Handles auth" (too vague)

Present tense. No marketing words. For the language to write summaries in, see "Summary language" below — never pick per-file.

---

Summary language

One run, ONE language. Every summary in the manifest — and the command's own narration — uses it. Without a single anchor, parallel sub-agents each guess and the manifest ends up mixing Chinese and English.

The command resolves the canonical language ONCE, up front, by this priority:

0. config.language from the todo output (set when .filetree.json pins language). When present it is authoritative — skip the rest of the chain.
1. Else the dominant natural language of CLAUDE.md / AGENTS.md (the agent contract — most authoritative).
2. Else README (any localized variant).
3. Else (/filetree:update only) the dominant language of existing manifest entries.
4. Else English.

Then it passes that one language verbatim into EVERY sub-agent prompt ("Write all summaries in <language>"). Sub-agents never re-detect; they run in parallel and would diverge if left to choose.

---

UNCHANGED bias (for /filetree:update ONLY)

Scope. This entire section applies to /filetree:update only. During /filetree:init the manifest starts empty, so there is no old summary to keep — UNCHANGED has nothing to refresh and apply will drop it. In init, every file gets a real summary. Do not apply this bias to init sub-tasks.

Why this matters. Hash changes trigger the LLM, but most code changes (typos, refactors, comments, small additions) don't change a file's purpose. Outputting "UNCHANGED" lets cmd_apply refresh just the hash and keep the existing summary — the manifest itself carries the memory of "I already reviewed this version". In a healthy update run, 80%+ of changed items should resolve to UNCHANGED. Writing a fresh 25-word summary when the old one still fits wastes ~100x more tokens than a 4-byte "UNCHANGED" reply.

Decision rule. You have: old summary, old hash, new hash, and the file's new content (prefer reading the git diff over the full file — diff is far denser per token and is all you need for purpose-level judgement). If the diff comes back EMPTY (the change was already committed, so working tree == HEAD), fall back to reading the file — the hash moved, so judging purpose from a blank diff would falsely yield UNCHANGED.

Output "UNCHANGED" if the old summary still describes the file's PURPOSE. Refactors, renames, bug fixes, test additions, formatting, comment changes, small additions — these almost always leave the purpose intact.

Output a new summary string only if:

• A major new feature has been added that meaningfully expands purpose
• A previously central concern has been removed
• The file has been substantially rewritten for a different goal
• The old summary is in the wrong language (not the run's canonical language — see "Summary language"). Rewrite it in the target language even if the purpose is unchanged; this is how a legacy mixed-language manifest converges — gradually, as each file's hash changes and re-enters the work plan. Language mismatch ALWAYS overrides the UNCHANGED bias.

When in doubt (and the language already matches), output UNCHANGED.

Rationalizations — every one resolves to UNCHANGED

The pressure to "be thorough" pushes toward rewriting. Each excuse below is a trap; the right answer is UNCHANGED.

Red flags — STOP, you're about to waste tokens

• About to write a summary that says the same thing as old_summary in new words
• Justifying a rewrite by how much the code changed rather than whether the role changed
• "Improving" or "tidying" a summary whose language already matches
• Reading the full file when the git diff already answers the purpose question

All of these mean: output "UNCHANGED".

---