diraAdd a reject disposition that tombstones an entry without freeing its id
Triage finding, 2026-09-02, from GH issues #35 and #36. Today the only way to remove a bad auto-capture (regex misfire, quoted duplicate of an existing decision) is to delete the file, which frees its number for reuse (#35) and which dira supersede refuses to touch because the target is state: staged, not confirmed (#36, with a misleading error: 'answer and 1 is never a verdict'). Proposed: a distill disposition (or a dira log/distill verb) that marks an entry rejected/tombstoned in place -- id retained, content marked disposed of -- so a junk capture can be cleared without either freeing its number or leaving an orphaned staged entry with no path to closure. Pair with a clearer supersede error naming the actual precondition ('entry state must be confirmed, not staged') instead of the current cryptic message.
Alternatives on record — 2
✗Make dira supersede accept a staged targetrefused supersede's meaning is already load-bearing elsewhere: docs/lore.md L-0003 and dec-0002 tie it to a decision genuinely being replaced by another decision, which is what dira check cites back at relitigation.
Overloading it to also mean 'discard a non-decision auto-capture' gives the verb two unrelated meanings the enforcer would have to distinguish.
✗Keep deletion as the only way to retire a staged entry, as todayrefused Deletion is the exact mechanism that frees the id for reuse -- issue #35's bug.
Keeping it as the only retirement path guarantees #35 and #36 keep recurring together.
revisit ifthe monotonic counter (dec-0032) ships and id reuse stops being possible even via deletion, at which point deletion may be safe to keep as the sole retirement path and this decision should be revisited