| name | align |
| description | One-shot semantic decision recorder. Writes topic + decision + rationale to memory.decisions for future sessions to recall. Tiny skill; load-bearing for cross-session continuity. |
/align — record a semantic decision
When you make a non-obvious architectural choice or pin a convention, run /align so future sessions know the decision was deliberate.
Inputs
topic — short slug (e.g., "session-tokens", "deps-policy")
decision — what was decided
rationale — why; ideally cites evidence
Output
Writes to memory.decisions (SQLite). Indexed by topic. Searchable via /memory.
When to use
- After a /council judgment.
- After picking between two viable approaches when the "why" matters.
- When deferring a future decision (record what was deferred and the trigger that should re-open it).
When NOT to use
- Trivial choices — taste.md is the better home for rules-of-thumb.
- Things that the code itself documents (the diff is the source of truth).