alignlisted

# Align **On the `model` field's enforcement (strategy-token-economy clarification 10, 2026-07-16):** `/align` is `user-invocable: true` — it runs on the interactive main loop, not via a `context: fork` subagent launch. A `model:` field in frontmatter is confirmed honored for `context: fork` skills; for `user-invocable` main-loop skills like this one, honoring is unconfirmed. This skill stays whole-session Opus because its interview dialectic is non-delegable and it has no dispatch launch path to carry an explicit `model: opus` argument the way `/align-tactics`' Step 3 subagent launch does. If the harness does not honor this field on the interactive path, the default here is intended-not-guaranteed — backed by the token audit's by-node/by-phase attribution (`strategy-token-economy`'s sensor) reading after the fact whether the session actually ran on Opus. `/align [optional requirement text]` records or revises a `strategy-*` intention node under interview. It supersedes `/file-issue`'s requirements-definition role for the graph-native dispatch model (`intentions/strategy-graph-native-dispatch.md`): a strategy enters execution by being recorded here, not by becoming a GitHub issue. Full spec and coverage matrix: `intentions/tactic-graph-native-dispatch.md` §2.2 and §4. This skill never files, edits, or closes anything on GitHub. Its only artifact is one or more `intentions/strategy-*.md` (and, incidentally, `intentions/tactic-*.md` draft) nodes landed on `origin/main`. **R