Supply the required directory collaborator to current runtime fixtures, verify absolute result navigation and deployment fallback, and retain the valid V4 working-directory recording instead of the unreleased V3 input that contains a V4-only event.
Delete @deepseek-ai/dsh-invariants, every package's ./invariant companion
and its tests, the invariant gates and Vitest setup, the scoped-event
resolver generator, and the INVARIANT-coded listener rethrow paths.
Update docs, READMEs, Agent Notes, and add an upgrade guide.
Review round 1: the settle assertions build their message only when one
is about to fail, so a passing run never pays the host-facts shell probe;
send windows are bounded by event sequence numbers instead of
milliseconds, so two sends that hand over within one millisecond keep
their own events; the replay starts at the send's own input write and
ignores the emulator's terminal-protocol replies; a throwing foreground
inspection is recorded before the send rejects; a sanitizer is attributed
to the most recently spawned terminal, which the class documents as the
sequential-session assumption. A unit spec pins the replay rule against
the session's evidence states, the window boundaries, the reply/input
distinction, and the bounded, escaped rendering. Notes and comments drop
change narration and the unverified claim about the pool's build.
The self-hosted Windows lane settles one or two of seven sends on the
controlled prompt and the rest at the plain silence bound, and the
settle reasons alone cannot say whether the marker was missing, its tail
was invalidated by later output, or the foreground comparison failed.
The case now records every decoded pty chunk with the sanitizer's marker
verdict and prompt tail, every foreground poll, every input write, each
send's settle reason and elapsed time, the session's prompt evidence
replayed from those chunks, and the host, pwsh, PSReadLine, and console
host versions; the settle assertions carry that timeline in their
failure message and onTestFailed prints it when the runner's budget ends
the case first.
promptTailGraceMs returns to the product default: master runs with and
without it took the same 20.8 s, so the lane's degraded sends never held
the state it covers, and a nonzero value there would only lengthen a
never-arriving-tail fallback.
Review round 1: the tolerance now requires the accumulated tail to remain a
prefix of the controlled prompt, so a prompt that arrived and was invalidated by
later output keeps the plain silence bound instead of waiting a tolerance with
nothing left to wait for. The field is validated as zero or at least one
`pollIntervalMs`, matching the handoff grace rule, and tests pin the extended
bound's expiry, the invalidated tail, and that rule. The owning readiness design
note records the new field and Tier 2 bound and cross-links this decision.
Windows sessions on the self-hosted lane settled one or two of seven sends on
the controlled prompt and charged the silence tier for the rest (issue 2487).
The marker is written by the shell's own prompt function and the printable tail
by the same render, so a tail that arrives past `idleSilenceMs +
handoffGraceMs` means a stalled delivery, not an absent prompt.
`dsh-terminal-bash` gains a validated `promptTailGraceMs` bound (default 0,
which reproduces the previous bound exactly). While a send has seen the marker
and its printable tail has not arrived, the silence bound grows by it. The
pwsh Loader-composition case sets 5000 ms, the host class the bound is for, and
still fails when no marker appears at all.
`operations.spec.ts` read the profile's run record from a 100 ms timer armed
before the refused installation whose repair it observes; a loaded runner
reached the repair run after that timer and the case failed with `null`
against the recorded run. The observing child now stays in flight until the
record appears, and the bounded wait starts at the launcher hand-over rather
than at test start.
The loader-composition case counts `stdin_read` settlements but reported only
the count, so its Windows failures could not say which tier settled which
send. The recorded reasons now ride in the assertion message; the threshold
and the assertions are unchanged.
Add a path-check reminder to the bash and pwsh tool descriptions. pwsh
also warns against assigning automatic variables such as $HOME; bash
also asks agents to guard path variables with ${VAR:?}.
- State the contract precisely: the cut never introduces a lone high
surrogate, and an unpaired surrogate the input already carries is not
repaired.
- Document the new export in the package README: a character-budget preview
section naming the persistent bash and pwsh shells and str_replace_editor,
plus a Known Limitations bullet that no longer contradicts it.
- Drop the assertion restatement from the two shell test comments.
- Let the str-replace test helper reuse a root so the clipped context reads
its file inside its own working directory.
Delete constraints the model learns from the call result, shorten
implementation detail, and move per-parameter rules into the owning
parameter descriptions. Bash, pwsh, and fs share one
sandbox_permissions description in dsh-sandbox.
The standard preset's first-turn prompt drops from 7985 to 6551 tokens.
Compaction threshold snapshots lower their context windows by the
1548-token tool-schema estimate reduction so each trigger keeps its
position; their fixtures move to v4 generations.
* feat(approval): localize prompts and allow consent for one turn
* test(approval): update cast baseline and Web PTC prompt pin
* fix(approval): cover PTC language and clarify turn consent
* fix(approval): group turn consent under allow menu
* fix(approval): keep turn consent in memory only
* test(snapshot): refresh owners for merged justification guidance
* fix(approval): drop duplicate allow-once from the allow menu
* refactor(approval): derive turn consent expiry from one check
Replace the turn approval bookkeeping (revocation state, incremental log
cursor, and turn-signal listeners) with one side-effect-free expiry check
over the log and lifetime flags. Events now only withdraw open
presentations early; the expiry check decides whether an answer or turn
consent still applies, including for Sessions without event publication.
* refactor(approval): drop one-turn consent and keep localization only
Remove the allowed-turn answer, the live turn-consent state, the allow
menu, and their tests, notes, and recorded Web scenario. The approval
service returns to the master request path; the request keeps the
localized displayReason for presentation, and the card resolves it in
the active UI language.
* test(snapshots): record the localized justification schema text in new scenarios
---------
Co-authored-by: Turtle <[email protected]>
Co-authored-by: Kaige-Gao <[email protected]>
Conflict resolution:
- docs/config-catalog and docs/module-graph: regenerate with the current
generators, which write the bilingual pages together.
- The conflict came from master's new short-fingerprint pairing records: our
branch still carried the previous full-hash records. Take master's record
format, then re-record the output-retention README pair after regenerating
its two sides.
* feat(i18n): key translation pairing records by heading section
Replace the whole-file blob hashes in *.i18n.yaml with one entry per
heading section that still holds language-specific content, keyed by the
English heading-slug path. Blocks identical across the pair (generated
table rows, code blocks, shared regions) stay out of the hash, so edits to
different sections and identical regenerations merge with Git's default
text merge, including on GitHub.
Remove the dsh-translation-pairing merge driver, its resolver command and
installer wiring; gen-translation-brief recovers last-confirmed text from
commit history. Migrate all active records, archive the superseded Agent
Note, and record the replayed conflict reduction in the new one.
* test(i18n): refresh translation prompt snapshot for the terminology note
* feat(i18n): exclude only code blocks and generated regions from pairing records
Replace byte-equality masking with an explicit rule: a record hash skips
fenced code blocks and BEGIN/END GENERATED regions, the two kinds of
content the pairing gate already requires to be identical on both sides.
Move generated data into regions written into both pages:
gen-config-catalog renders both languages and the record with per-package
regions labelled by code identifiers; gen-module-graph keeps its table in a
region; gen-doc-graphs splices the event matrix regions into the Chinese
page. Add a generic region splice beside the region grammar.
Conflict resolution:
- docs/config-catalog: regenerate the English page; the Chinese page follows the
three shifted Config sources and the bilingual pair record is re-recorded.
- docs/module-graph: the generator rewrites all three artifacts together.
- packages/shell/tool-pwsh-persistent/src/index.ts: keep master's rewrite of the
prompt handling and re-apply the surrogate cap and its import.
- The three package.json dependency entries follow master's workspace:* range
convention, and the lockfile importer specifiers match them.
* feat(tool-cordis): inspect live plugin Config schemas
Replace the Host `Builtin` inspect provider, which described a dynamic Host
half the model can no longer define, with a `Config` provider that lists every
live Loader entry with its Config status and projects one entry's native
Schemastery Config into a self-contained JSON Schema document. app-boot exports
its projector, native-schema check, and the shared `loaderExpression`
definition for the live walk; the CLI dump keeps its profile-composition path.
* fix(tool-cordis): refresh config catalog pointer and presets e2e provider list
* feat(agent-preset): progressive creator skills and profile shell facts
Split `cordis-plugin-development` into a short procedure with `references/`
read on demand and copyable `templates/`, replace the source-reading ban in
both creator skills with an ordered list of knowledge sources, and add the
`cordis-composition-reference` skill with the Loader patch dialect and a
generated, doc-sync-gated list of loadable plugin packages. `shell-env`
exposes `DSH_PROFILE` and `DSH_PROFILE_DIR` so skills can address the installed
profile.
* chore(agent-preset): regenerate plugin package list after merging master
* fix(tool-cordis): resolve the Loader at Config query time
A Host composed programmatically without a Loader keeps registering the
inspect providers; `Config.listConfigs` reports the missing Loader when
queried. Refresh the advanced-toolchain recordings whose provider list now
carries `Config` in place of `Builtin`.
* chore(agent-preset): regenerate plugin package list after merging master
* docs: regenerate module graph for the tool-cordis app-boot dependency
* docs: regenerate module graph for the shell-env app-boot dependency
* docs: regenerate event graph pointer for app-boot exports
* fix(agent-preset): address review on creator skills
`Config.listConfigs` reports each entry's resolved `packageDir` through the
profile package lookup, and the skills read READMEs and built declarations
there instead of guessing `$DSH_PROFILE_DIR/node_modules`, which holds only
profile-installed packages. Skills name the PowerShell variable syntax and the
no-profile omission; the composition reference no longer attributes
`overridden` to `list_plugins`; references point at the templates instead of
pasting them. The skills spec renders every skill and bounds it by the pruner
threshold in code points, checks referenced files and the source-reading ban
across all skill files.
* fix(tool-cordis): address review on the Config inspect provider
Report the Loader entry id (unique across include trees) beside the tree-local
patch id, classify group and include carriers as `tree`, treat disabled and
disposed fibers as `inactive`, and drop the unreachable boolean-schema branch
by narrowing the projector's root type. The presets e2e now queries
`Config.listConfigs` over the booted Web profile and projects the shipped
`tools` row; the advanced-toolchain recordings carry the updated manifest.
* fix(tool-cordis): page the Config directory and filter it by plugin name
A full directory of a shipped Web profile exceeds the 50000-byte spill cap
once rows carry more fields, and the spilled preview is no longer JSON.
`Config.listConfigs` now returns pages (`offset`, `limit` up to 100, default
25) with `total` and `nextOffset`, and an exact plugin `name` filter; the
presets e2e reads the shipped rows through the filter.
* chore(agent-preset): refresh skill recordings after the paged Config query
* docs(agent-preset): teach the composition skill to migrate legacy presets
A preset authored before declaration rows lived in
`$DSH_HOME/.agent-presets/<id>/` as `preset.yml` plus `agent.cordis.yml`;
the skill now maps those files onto a declaration bundle and re-checks plugin
names against the package list.
* docs(agent-preset): read skill files through the file-read tool only
Desktop ships the skill directory inside app.asar, which the Host process can
open but shell commands, node, and pnpm cannot. The skills now say to read
references and templates with the file-read tool and copy templates into the
workspace before installing or checking them; the spec rejects shell reads of
skill files.
* docs(agent-preset): document bundle display metadata and package.json icon in the host-plugin reference
Three tools capped the preview text they return with String.prototype.slice
at a UTF-16 code-unit budget. A cut that lands inside a surrogate pair
leaves a lone high surrogate, which is not well-formed text: a strict JSON
reader rejects the durable Session log carrying it, and the preview shows a
broken glyph. A lone surrogate retained in history has already blocked every
later Messages request for one session.
Centralize the cap in output-retention as
truncateWithoutSplittingSurrogatePair, and use it from the persistent bash
and pwsh shells and the string-replace editor. The cut drops the unpaired
half instead of keeping it, so the retained text stays a prefix and is at
most one code unit shorter than before.
The new dependency edges and the shifted Config sources regenerate
docs/module-graph and docs/config-catalog, including the committed
translation pair.
Refs #4817
The composition suite's idleSilenceMs: 30000 was below the suite's own send
bound (dsh-terminal-bash timeoutMs: 300000), so the comment claiming silence
could settle nothing was wrong: a regressed send still settled as inferred_idle
after ~30 s and only the accumulated case timeout caught it.
The case now records every send's waitReason and requires at least six
stdin_read settlements with no inferred_idle. Reinstating the removed prompt
override fails it in 30 s with 'expected 0 to be greater than or equal to 6'
while the output assertions still pass. The silence tier returns to its product
default.
Also renames the two stale test titles ('private shell prompt', and the
cancellation phase that no longer covers an initialization send), pins the
stdin_read fallback's session retention, and records the interactive-child
consequence of that retention in the package README pair and the Agent Note.
tool-pwsh-persistent installs its own `prompt` function after spawn, so
terminal-bash prompt readiness never matched and every send degraded to the
silence tier (idleSilenceMs + handoffGraceMs) under production defaults. This
is the defect the persistent bash tool fixed for the bash stack.
The tool now initializes without a prompt send and detects the no-end-marker
fallback through the seam's stdin_read wait reason instead of matching its own
prompt text. The private prompt constant and its stripping are deleted, so
partial output can carry the backend's own trailing prompt.
Tool calls drop from 8493/3722/3832/3759 ms to 1340/255/251/241 ms (spawn and
initialization plus the first command, then three warm commands; Windows,
production defaults, real Loader composition). The loader composition suite
pushes idleSilenceMs beyond the send bound, so every case fails if prompt
readiness regresses.
With a job registry composed, tool-bash and tool-pwsh now register every
command with ctx.jobs as it starts and wait on the job with the registry's
own bounded wait: a command that finishes in time returns the ordinary
foreground result and its record is removed, one that outlives the timeout
returns the same job id it already had, and a kill from outside the call is
reported as `[stopped: <reason>]`. Nothing changes hands at the deadline, so
the shell seam loses `onExpiry: 'offer'`, `ShellExecution.promotion`, and
`ShellPromotionOffer`, and both local executors drop their offer arm.
The registry stays optional: the tools register a foreground-only definition
while no registry is composed and swap to the job-backed one inside a
`ctx.inject(['jobs'])` fork, so a composition without jobs keeps a plain bash
with no job tools in the model's schema. A refused registration at the start
(admission, no controller) runs under the executor's deadline kill, and a wait
that expires before the process spawned ends as the preparation timeout.
JobRegistry gains `remove(id, caller?)` and the `settled` event reports
`awaited`, so dsh-tool-jobs no longer keeps its own wait ledger; it withholds
notices for awaited settlements and for kills the model requested. The tool
catalog harvest composes the registry so the shipped schema stays documented,
and the two compaction fixtures move their context window from 10000 to 10300
tokens because the branch's larger tool schemas pushed their condensed history
over the old trigger, which requested a fourth scripted model call.
Tianyi Cui