--- name: loom-mentor description: > Use when a loom hint (a tool result's `hint`, an `X-Loom-Hint` header or a "Loom mentor" line) names a plane or op you do not know how to use, or when the user asks "what can loom do", "how should I be using loom", "am I using loom well", "loom mentor". A hint that already names the op to call needs no skill: call it or carry on with the task. Maps a situation to the loom plane and op that fits — code index, tracker record, documents, records, working memory, offload, the entity graph, runs, decisions, procedures, duplicates, conflicts, bisect — and says what to do about each hint code. NOT for deciding the shape of one write (loom-memory), NOT for grep-vs-code.* in an indexed repo (loom-code), NOT for reworking a namespace that already holds the wrong shape (loom-optimize). disable-model-invocation: true --- # loom-mentor — the store's advice, and where each plane lives Loom has a few hundred operations; your tool list shows the ones a coding session uses most. The mentor reads what your agent called this week and names the plane it is not using: a `hint` on a tool result, a `Loom mentor` line in the session primer. Read a hint once, act on it once; the same code is not shown again this session. This skill routes. The destination skill owns the detail: the shape of a write is `loom-memory`, code questions are `loom-code`, reworking stored rows is `loom-optimize`. ## Calling an op your tool list does not show A listed op is the tool `loom_` (`tracker.similar` → `loom_tracker_similar`). An op that is not in your tool list runs through `loom_op {op: "tracker.similar", arguments: {…}}`; `loom_op {op: "help", query: ""}` lists matching ops and their arguments. Your host can only call tools it was shown, so never call an unlisted `loom_*` name directly. The tracker tools are listed when the MCP entry mounts `/mcp?toolset=coding+tracker` (the plugin's `toolset` option), or a second entry mounts `/mcp?toolset=tracker`. ## The planes a hint points at | Situation | Op | |---|---| | "has this crash / test failure / report been seen before" | `loom_tracker_duplicate_check` (listed: draft title + body → `duplicate_of` / `related` / `none`); `loom_tracker_similar`, `tracker.flaky`, `tracker.test_signal` (listed on `coding+tracker`, else through `loom_op`) | | which PR broke a behaviour between two versions | `loom_code_bisect_candidates` (symbols or paths, `good_version`, `bad_version`) | | about to open a PR | `loom_code_pr_check` (title, body, paths, symbols, issues) | | is it safe to change X | `loom_code_impact` | | a document, runbook or report over ~2 KB | `knowledge.upsert` + one pointer row (`loom-memory`) | | a table of facts with identity (PRs, alerts, services) | `memory_schema.register` once, then `memory.record.write` | | your plan, a lease, "where was I" | `loom_working_memory_set` | | a 3,000-line log or diff to park | `context.offload`, then `context.fetch` | Tracker ops take the tracker namespace, or the repository's code namespace when it is bound to one (`tracker.binding` says which). ## The hints you will see most | code | do | |---|---| | `tracker_unused` | run the tracker check above before filing or grepping | | `code_index_unused` | `loom_code_symbol` / `loom_code_search` for identifier questions | | `bisect_available` | `loom_code_bisect_candidates` first, blame second | | `get_miss` | `loom_search` by meaning; `memory.list prefix=`; `follow: true` | | `set_duplicate` / `duplicates_written` | revise the named row, never a third copy | | `documents_as_memories` / `set_large_value` | a knowledge document plus a pointer row | Every other code, the full situation table and `mentor.brief` (the counts behind a hint) are in `reference.md` beside this file. A code that is in neither table is newer than this skill: follow the hint's own sentence and the ops it names, or ask `mentor.brief` for its evidence.