feat(tui/subagent): resume interrupted children from checkpoint via followup

[SparkofSpike]

Summary

agents/followup on an interrupted_continuable child used to queue a dead-letter: the checkpoint was preserved and its continuation_handle returned, but nothing could actually resume the run — long tasks (document review, multi-step search) interrupted mid-way had to be re-dispatched from scratch, wasting tokens and time. This change wires checkpoint-based continuation: followup now re-dispatches a fresh agent loop seeded with the original prompt plus the checkpoint message tail and the follow-up text, under a new agent id, while the prior terminal record stays immutable.

Process: necessity review (scout) confirmed no resume substrate existed on main; implementation went through two adversarial review rounds (3 majors fixed: duplicate-resume idempotency, cwd-claim double-namespacing, runtime-posture widening).

Changes

• resume_from_checkpoint (SubAgentManager, mod.rs): re-dispatch via spawn_background_with_assignment_options with the interrupted child's agent_type/assignment/allowed_tools/model restored.
  • Idempotency: resume_targets map (old id → resumed id) — a repeated followup returns the existing target and forwards the message to it (never silently dropped).
  • Spawn-seam fidelity: child runtime via background_runtime() (depth +1, detached cancellation), depth ceiling checked before derivation; workspace restored to the interrupted child's directory; write claim restored from the coordination ledger (pre-namespaced, isolation flag preserved); runtime posture (read_only / denied tools / shell) preserved from the worker record instead of rebuilt from the caller's role.
• checkpoint_messages_to_text / build_resume_prompt: serialize the checkpoint message tail into a resume prompt ([RESUMED SESSION — checkpoint …] + prior conversation tail + parent follow-up).
• AgentsFollowupTool (coord.rs): optional SubAgentRuntime via with_runtime; execute routes interrupted_continuable children with a runtime attached to the resume path, keeping honest queue-only semantics otherwise. Both registration points (AgentTool followup action, register_coordination_tools) attach the runtime.
• Tests: 5 manager tests (seeded resume, workspace restore, idempotency with message forwarding, reject non-interrupted, reject missing continuable checkpoint) + 2 coord tests (queue-only without runtime, resume with runtime). stub_runtime made pub(crate).

Type of Change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Refactoring (no functional changes)
• Tests

Testing

• cargo fmt --all -- --check
• cargo clippy -p codewhale-tui --bin codewhale-tui (no new warnings; only the pre-existing mcp.rs:262 warning inside a #[cfg(windows)] block, invisible to the ubuntu CI runner)
• cargo clippy --workspace --all-targets --all-features --locked (warning-free under the CI allow list) — not run locally; CI will run it
• cargo test -p codewhale-tui resume_from_checkpoint (4/4 pass) and cargo test -p codewhale-tui followup_interrupted_continuable (2/2 pass)
• cargo test -p codewhale-tui tools::subagent (392/392 pass)
• cargo test --workspace --all-features --locked — not run locally (heavy); CI will run it

Checklist

• Updated docs or comments as needed
• Added or updated tests where relevant
• Verified TUI behavior manually if UI changes (N/A — no UI change; resume path exercised via unit/coord tests)
• Harvested/co-authored credit uses a GitHub numeric noreply address (no human co-authors; agent assistance noted in commit body)

Related Issues

No-Issue: harness improvement backlog item (sub-agent checkpoint resume), no public issue number. Necessity-review and two review rounds performed as part of the implementation workflow.

[Hmbown]

Reviewed against the v0.9.4 candidate: checkpoint-resume is squarely in the child-lifecycle scope the candidate is hardening (persistent subagent visibility + recovery). When this leaves draft, we'll review it on the candidate branch — please keep the checkpoint format compatible with the current subagent checkpoint records.