[shipmill-romamo]

Spec for #26 (its last open part, stale, back for a decision after 0.6.0 shipped stable).

docs/specs/001-stale-proposals.md, status: approved. Merging this PR approves the spec; no build starts before then.

What the spec settles, for review:

• Rule: a proposed decision is stale when a host its basis.observedAt names now has a later observedAt in the basis folder. The observation timestamp only: the basis records no repository commit, so a commit rule needs the basis to grow first (out of scope)
• Derived, not stored: staleness is worked out on read. No schema enum change, no write on read, no sweep command, and a record still loads in an older cloudfall. This departs from D-1's pattern of a stored status, on purpose: a stored stale would need a trigger that writes records
• Surface: stale and staleHosts on proposed decisions in decisions list, decisions show and why; a --stale filter on decisions list and why
• Approve doesn't refuse: with timer-driven observations (the M10 pattern) every proposal goes stale within minutes, so a refusal would block them all. The attestation prompt names the stale hosts instead

specs.py check passes.

🤖 Generated with Claude Code

[shipmill-romamo]

@romamo Decision needed: approve spec S-001 (stale proposals, #26)?

Merging this PR is the approval, so triage won't merge it under "merge when green". The choices it makes, for you to confirm or change:

1. Merge as is (recommended): staleness by observation time only, derived on read (no schema change), reported in decisions list / show / why, --stale filter, approve prompt names stale hosts but doesn't refuse
2. Same, but decisions approve refuses a stale proposal with PRECONDITION until it's proposed again. Safer, but with timer-driven observations every proposal goes stale within minutes
3. Store stale as a status value (D-1's pattern) with a sweep command that writes it. Filterable with --status stale, but a new enum value an older cloudfall refuses to load

Merge for 1, or reply with a number or your own answer; shipmill takes this up on the tick after your reply.

[shipmill-romamo]

Decision by @romamo, relayed by Claude Code

1, merge as is. Roman merges this PR himself.