skills/legalquants/references/endings.md

# The endings table — every companion skill's door

Canonical copy for PRD-02. Each SKILL.md carries its own ending in place; this
file is the audit source. Every ending: one breath (what's finished → what's
saved or "nothing saved" → the door), declinable, never a command run for the
user, and the disclaimer line always last.

| Skill | Door | Rule |
|---|---|---|
| `lq-apply` | assess.legalquants.com | The application is LQ Assess (the assessment is the application); named once, when they are ready. The public profile lives on legalquants.com once real. |
| `lq-ask` | legalquants.substack.com | Name the sources used, then the honest scope: the paid Substack carries the full archive and member-discussion summaries. |
| `lq-reflect` | `$lq-connect` | Only when the session surfaced something unresolved and human. Nothing stuck → no door. Never invent a stuck. |
| `lq-mirror` | first skill to try | One legal-workflow skill from the live catalog output, never from memory, with its `$` command and why it fits. |
| `my-lq-moment` | `$lq-apply` | Earned moments only: this moment is evidence; apply is where it compounds. An honest-no session ends with its pointer, no door. |
| `lq-connect` | none | It is the destination. |
| `lq-start` | none | Its closing line ("Tell me what you're working on…") is the routing close, not a door. |
| legal-workflow skills | none | Work tools end with the work receipt. Fenced — never add a funnel to them. |

No skill launches another, contacts anyone, or saves anything as part of an
ending. The only write path anywhere is the store contract's consent-gated
`save`, and it always runs before the ending.