Semantic resolutions beyond line merges:
- The bash seam keeps resolveMode + the bash/resolve-mode waterfall on
master's task-free BashExecutor (run/start/resolve only; BashProcess
handles); tool-bash consults it at its stamping site and escalation
baseline on master's render/background split, with a waterfall test on
the recording executor.
- dsh-mode's BASH_FAMILY narrows to ['bash']: bash_output/bash_kill are
replaced by the kind-generic task_output/task_kill, which span every
task kind and only observe or stop work, so the access cap withholds
only the starter it can reason about.
- The plan-mode snapshot suite adopts master's pin grammar (tool-schema
sidecars; the expectedHeaderSnapshots extension is gone — the exit
transition deltas, and entering-before-turn-1 needs no second
snapshot); modes-advertise joins the plan header class (no-model, so
membership is vacuous). Fixtures re-recorded on the acp-demo bin;
the replay overlay gains the passthrough sandbox runner.
- examples/plan-acp-agent rewires to @deepseek-ai/dsh-acp-demo and drops
its tool-bash entry (the spine bundle now composes it); dsh-stdio (the
renamed stdio-chat home) keeps its /mode command and gains the dsh-mode
peer edge; the acp bridge keeps the modes surface beside master's
permission presets.
- mode README gains the Model Experience / Known Limitations sections the
new README gates require; AGENTS.md ceiling 1370 → 1440 for the kept
mode/ layout line and Agent efficiency section.
The public kill result used the awkward phrase already-terminal. Rename it to already-finished and keep the model-facing response aligned; not-alive would be inaccurate because a force-failed registry record can still correspond to orphaned producer work.
Task kinds were open strings even though producer namespaces are an extension point. Add the merge-extensible TaskKindMap and derived TaskKind, cover consumer declarations in task and bundle tests, and retain the runtime non-empty check for untyped callers.
With exactOptionalPropertyTypes, owner?: Agent | undefined allowed an explicit undefined value that no caller needs. Tighten the property to owner?: Agent so unowned work is expressed by omitting it.
Record the requested task-service/backend split as a follow-up, using a systemd-backed runtime as a concrete candidate without guessing its durability and ownership contract in this PR. Regenerate the type and Cordis catalogs so public docs match the declarations.
Comparing a feedback-time PR commit directly with the landing merge lets unrelated target-branch changes satisfy a review comment and creates false adoption evidence.
Define feedback-time and final PR-specific patch snapshots, fail closed when the landing shape cannot be reconstructed, and require a target-only negative control before implementation.
A saved complete-file candidate could overwrite newer skill guidance, while a provider-wide outage was indistinguishable from a healthy no-op run. Record the source blob and provenance, reject drift during promotion, and make total adapter failure visible.
Align cadence and feedback acquisition with the evidence contract, and narrow the borrowed-state and negative-control checks so they do not contradict the package and testing standards.
Refresh the background-task branch onto the latest repository baseline before applying review feedback. This preserves merge ancestry, incorporates the current CI and package-structure changes, and ensures the review fixes are validated against the code that will actually receive the PR.
Providers stream token-sized deltas, so a session log stores hundreds of
near-identical assistant/chunk lines whose JSON envelopes dwarf their
payloads (~56x measured on a real DeepSeek session, 73% of file bytes).
Add a lossless storage codec to dsh-session: packChunkRuns() folds each
run of >=3 consecutive same-block delta chunks into one storage row --
text-chunks / reasoning-chunks / tool-call-chunks, bare slash-less tags
like the header line's 'session' so rows cannot be confused with session
events -- and decodeStorageRecord() expands rows back to the exact
original events (seq0/time0 + dt gap array reconstruct every member's
seq/time; tool-call rows carry the run-constant id/name). The encoder
whitelists exact shapes and stores anything unrecognized verbatim; the
decoder validates row-tagged values and fails loud on malformation.
The JSONL backend gains a packChunks config (default false). Writing
packs only when enabled -- default-off output stays byte-identical to
the previous layout, so snapshot goldens are untouched. Reading is
layout-blind: scanLog always decodes rows and now checks seq contiguity
with a cursor instead of the line index, so packed, unpacked, and mixed
files all load identically. Fixture readers (llm-replay parseSessionLog,
acp-snapshot normalizeSessionLog) share the codec; the normalizer zeroes
a row's time0/dt exactly like an event's time. The two demo bundles
plumb packChunks from cordis.yml to the backend.
Measured on a real coding session: 105 KB -> 42 KB (-60%), 475 lines ->
74, with reasoning/tool-call heavy sessions saving the most. Covered by
example + fast-check round-trip codec tests, backend packed/mixed/torn-
tail specs, and an end-to-end demo run loading a packed log through a
default-config backend.
The background-task change repeated its lifecycle design across implemented RFCs, package READMEs, JSDoc, test commentary, and model-visible schemas. That repetition obscured the contracts that maintainers must preserve and added avoidable prompt tokens.
Rewrite the implemented RFCs around the current design, keep authorization, exact-owner cleanup, wait/abort ordering, producer quiescence, and teardown-failure guarantees at their owning surfaces, and remove peer surveys, review history, control-flow narration, and emphatic restatement.
Shorten the task and subagent schema wording, synchronize the bilingual tool cookbook, and regenerate the config, service, RFC, tool, and replay snapshot derivatives. Runtime behavior is unchanged; test edits update prose-only assertions and descriptions.
Entry point for the skill's operator and for any future handoff: what the
private periodic tool produces, where the saved candidate diff and SKILL.md
land on the operator's machine, and how the promote helper turns the
operator's decision into a draft PR from a clean master checkout. Companion
to the maintenance RFC.
A private out-of-repo maintainer tool turns adopted human review feedback
into candidate skill updates. Only the resulting SKILL.md diff enters this
repository through normal PR review; the tool source, adapter binaries,
provider credentials, and scheduler stay on the maintainer's machine to
keep this repository focused on shipped product code. Kept in proposed/
until end-to-end verification against master lands a candidate PR; the
acceptance criteria list the exact observations that unlock promotion.
Tianyi Cui
imccyu
kingwl
Yichen Jiang
NI0317