fin

name	fin
description	/fin — drive the current goal to completion, no confirmation stops. NOT for a single command (just run it), NOT for resuming interrupted/paused agents or tasks from earlier in the session (use con, the multi-goal recall mode).
when_to_use	finish this, keep going, don't stop, just do it
user-invocable	true

/fin — finish mode (the current goal, to done)

Drive the current goal to completion. Don't stop until it's actually done.

Behavior

1. No stopping early — Continue working through all pending tasks
2. No unnecessary questions — Only ask if results are clearly ambiguous or conflicting
3. Verify completion — Check all tasks are done before finishing
4. Self-correct on blockers — If something fails, find new ways. Grind harder before asking. Verification fails twice → try a third angle, not "should pass".

When invoked

• Review current task list (if any)
• Execute all pending work
• Verify each task completed successfully
• Only stop when nothing remains or blocked by true ambiguity

What counts as "ambiguous/conflicting"

• Two valid approaches with significant tradeoffs
• Contradictory requirements in the request
• Missing critical information that can't be inferred

What doesn't count

• Implementation details (pick one and go)
• Minor style choices (use existing patterns)
• Uncertainty about best approach (try the obvious one first)

Stopping discipline (the hard part)

Before declaring done, run an open-items pass:

1. Re-read the most recent diary entry's open-items list. Anything tagged [open <date>] or in a "open items for next session" list is a candidate.
2. Scan your last 5 outputs in this conversation for any of: "I'll … next", "still TBD", "pending", "next:", "next sprint", "queued", "leftover", "deferred". If you said it, it counts.
3. List in-flight subagents you spawned and haven't received completion for. They are pending work.

For each candidate item, classify:

• Active (this session) — finish it now. The user asked you to /fin; that means resolve.
• Deferred (next session, with reason) — explicit log: "deferred because ". Diary entry must reflect it.
• Blocked (waiting on user) — ask the user a focused question; do NOT silently stop.

Never declare done with active items unresolved. "Tree clean + no in-flight subs" is NOT a stopping criterion. The criterion is: every item from the open-items pass is in Deferred or Blocked, none in Active.

Cascade guard

If resolving an open item creates more than 3 new open items, stop and report the cascade to the user before going deeper. Otherwise /fin can spiral.

Common anti-patterns

• "Natural stopping point" — no such thing in /fin. Either done or deferred-with-reason.
• "I have nothing in flight" — irrelevant; check the open-items list.
• "I'll come back to that" — fine if logged as Deferred with a reason. Not fine as a silent assumption.