Verify queued forwarding retains exact unread mail on cancellation before start. Keep unavailable stores unknown, exclude inputs and bookkeeping, preserve split actor stores and delivery identity. Document host factual speech.
Timer follow-ups carry the same objective_author stamp a promote carries (B3). The host_task_facts row and the cancel receipt state files_rescued as a stat-only positive/zero/unknown count with hash_computed:false, walking the canonical and child-drive artifact stores (C2 fallback per owner correction; B5 receipt stays NOT_DONE pending the TZ-1 landed API). New real-server Chromium test for a zero-option question card (B1). Chapter budgets bumped with rationale.
Co-authored-by: Ouroboros <311266734+ouroboros-agent@users.noreply.github.com>
Keep this version-neutral contribution local. B1/B2/B3 and C1/C4/C5 are partial; C2/C3, exact diff review, full suite and PR are outstanding. Do not merge this checkpoint without final-candidate checks.
Add caller-owned Claudexor model calls, unified subscription/API setup, explicit role accounts and live quota/auth continuation. Preserve prompt/tool ownership, physical-attempt custody and manual assignments. Keep the existing task process through waits and post-work.
The reviewed source checkpoint retains the current dependency pin. Published signed Claudexor bytes, cross-platform CI and the agreed merge ordering remain delivery prerequisites. No Ouroboros version, tag or release is created.
Co-authored-by: Ouroboros <311266734+ouroboros-agent@users.noreply.github.com>
Round-2 reviewer finding (NOT-CLOSED): the producer cutover left stored
legacy aliases flowing OUTBOUND through generic passthrough — the
TASK_COST_META_FIELDS carry list still named both spellings, write_task_result
re-persisted **existing aliases, public_task_result deep-copied them into the
task detail, the history mapper copied both spellings out of stored rows, the
cancel path relayed them via the meta-field loop, and the fan-out sweep's
(file, alias) allowlist let a new emission in an allowlisted file pass.
Projection-boundary semantics now landed:
- The ABI carries NO alias: public_task_result normalizes the top level plus
the two nested planes that reach the payload (subagent envelope,
loop-outcome usage snapshot); task detail and the list row inherit it.
- Stored legacy reads deprecated-wins and NORMALIZES at re-write:
write_task_result merges over with_cost_aliases(existing) and normalizes
the merged row (a legacy mutator's own edit still wins its pair, then is
stripped).
- TASK_COST_META_FIELDS is honest-only; every consumer that copied it from a
possibly-legacy source converts via carry_cost_meta instead (history
task-summary replay, progress-meta replay, terminal-truth annotate, cancel
stored/child cost, task_cost_finalized event). post_task_checkpoint keeps
an explicit SCRUB set that still pops the retired spellings so a stale
legacy replica cannot smuggle an amount past deprecated-wins at the write
seam.
- Producers whose data reaches the public projection are cut over and
REMOVED from the sweep allowlist: build_subagent_envelope/envelope_from_task
stamp accounted_upper_bound_usd (kwarg renamed), the pipeline
unavailable-patch, and the loop-outcome usage sub-dict (extracted to
module-level _loop_usage_snapshot in the same edit — the honest-name
comment had pushed derive_loop_outcome past the 300-line law; round-1
entry-7 precedent, landed INSIDE this commit).
- Fan-out sweep upgraded: keyword args on ANY call are emission-shaped now;
the allowlist is PER-SITE (file, alias, enclosing scope) with staleness
failing the test; a dedicated pin bans allowlisting the public-projection
planes. Generic passthrough is covered by the new runtime
projection-boundary pins (stored legacy row -> public_task_result / task
detail / list row contain only honest names, deep-scanned; rewrite
normalization; legacy-mutator honor-then-strip).
Test-contract disclosure (OLD-ABI clauses converted to honest names, per the
round-2 mandate): test_gateway_history (windowed anchor, terminal cost truth,
override precedence, nullable bounds, task-summary passthrough),
test_tasks_list_slice compact-row cost clause, test_cost_projection meta-field
derivation + carry tolerance, test_task_summary snapshot fixture (stale legacy
key; the real producer emits the honest name), test_headless_task_artifacts
mirror-cost merge + finalized-accounting clauses, test_task_result_monotonic
kept-cost clause.
(cherry picked from f1-d01 commit 1039baed and squashed with the seam so the
manifest is self-consistent: the pick-time --theirs manifest predated the
D10/D11 hunks, and a follow-up regen would have left an inconsistent
intermediate commit - the 7d2dca49 class; recipe fixed to regen BEFORE
cherry-pick --continue.)
D01: loop.py 7003->865 + nine leaves, agent_dispatch + post_task_synthesis,
177 spans proof-green, zero live v7 deltas in the loop split (55 falsified
spans are all upstream drift), maximal declared sets per the D10 lesson,
HOT_CODE_PATHS closed over the nine loop leaves.
Seam: D11 six server leaves and D10 nine git leaves mapped (D01 mapped its
own eleven in-lane); server.py, tools/git.py, supervisor/git_ops.py split
rows retired - the remaining 21 split_pending rows are exactly the F2
hot-organ set. size_ratchet manifest regenerated by the official tool on the
merged tree. test_module_handle_extraction LEAVES = union of D10 and D01 row
sets (restored the git_ops_updates closing bracket the conflict marker had
split; 78 passed). Quotient report regenerated: 1426 strict module edges.
