Six review findings on the select surface, all reachable from the wire:
**Resume read the header, not the log.** The switch was recorded as
`agent-preset/selected` and every projection resolved from it, but `agentFor`
still composed from `inspected.meta.agentPreset` — the value written once at
creation. A blank session that switched and then ran turns came back after a
restart under the ORIGINAL preset, restoring that history under the tool set it
was not produced with, which is the mismatch this feature exists to prevent.
`inspected` already carries the events.
**Cold summaries dropped the preset entirely.** `summarizeCold` hand-copied
three header fields and omitted the fourth, so a restored session reported no
preset and the picker showed the deployment default. It now uses the same
projection the attached path does.
**`select` had no gate.** Two concurrent selects both passed the blank check;
the second `unmountPresetFor` then found no record, because the first had
already removed it, and both mounts installed into one agent layer. Selects on
one session now queue, and the blank check is re-read inside the queue. This is
not turn admission — a `session.prompt` racing a switch is the agent loop's to
reserve — but it closes the select-versus-select tear-down.
**A same-id restore was skipped.** The roster is a live directory, so "the same
inputs that worked a moment ago" does not hold: a changed file is exactly how a
same-id reselect fails, and skipping the restore left the agent with no
composition at all.
**`writable` was dead state**, initialized true and never set, so the row could
never disable. It now carries `settings.describe`'s bit — a browser that may
not write settings sees the current default and no control, rather than one
whose write answers `settings-not-exposed`.
**`list` was documented as id-ordered.** It is root-precedence order with each
root's own presets sorted, first root to supply an id winning.
The inline comment called every dunder entry "the Python backend's" slot,
but `__debug__` is a CPython compile-time constant, not a seeded slot;
align the comment with the seam declaration, which is the single home for
why each entry is reserved.
- index.ts: RESERVED_ERROR_MEMBERS JSDoc now says `__x__` (non-empty
middle), matching DUNDER_MEMBER and the derivative docs; and
PORTABLE_RESERVED_WORDS is described as covering portable *target*
languages (Python is a later-PR backend, not yet shipped).
- types.ts + type-equiv docs: CodeBindingNamespace.global points to
RESERVED_BINDING_GLOBALS by name with an `e.g.` sample rather than
enumerating the set (single home), no longer implying `__debug__` is a
seeded slot.
- Agent Note: separate `__debug__` (CPython compile-time constant) from
the seeded bootstrap globals.
- README (en/zh): document the four exported exclusion sets and the
portable identifier/naming contract — no cross-package link needed.
- Regenerate cordis services catalog for the shifted source line.
- Agent Note: correct the stale worker-adoption paragraph — the worker
consumes the seam constants directly by name (no RESERVED_WORDS /
RESERVED_ERROR_PROPERTIES re-alias) — and describe DUNDER_MEMBER as
`__x__` (non-empty middle).
- types.ts: document RESERVED_BINDING_GLOBALS on CodeBindingNamespace.global
(names like `__dsh_main__` pass the identifier rule but are still refused)
and the non-empty-middle dunder rule on memberNameProperty; propagate to
the type-equiv derivative docs (both languages, re-recorded).
`____` has an empty middle between the two `__` pairs and does not match
`/^__.+__$/`; assert that (not a match) and add `__x__` as the shortest
real dunder form that does.
- Worker consumes PORTABLE_RESERVED_WORDS / RESERVED_ERROR_MEMBERS by
their seam names directly, dropping the local re-alias (symmetry with
the other two imported constants).
- Split the reserved-vs-duplicate diagnostics: a backend-owned global now
reports "reserved binding global", not the misleading "duplicate".
- DUNDER_MEMBER uses `__.+__` so a bare `__` (empty middle, not a real
CPython dunder) is not matched; add coverage.
- Worker misuse tests add `a$b` (second-char `$`) and `lambda` (Python
keyword) so the identifier narrowing and reserved-word adoption are
each pinned directly, not only transitively.
- Clarify the seam JSDoc (dunder-vs-explicit-set wording, Python backend
is a later stack PR) and record in the Agent Note the obligation to
widen RESERVED_BINDING_GLOBALS when the bootstrap seeds more globals.
Update the type-equiv blocks in docs/core-data-structures/code-runtime
(both languages, re-recorded) to mirror the new CodeBindingNamespace /
CodeBindingErrorClass JSDoc, and regenerate the cordis services catalog
for the shifted source line. Keeps the worker-only intro until the
Python backend package exists.
Move the reserved-word, reserved-global, reserved-error-member, and
dunder exclusion sets from the worker backend up to the code-runtime
seam package, and narrow the portable identifier subset to drop the
JS-only `$`. Every backend now imports one contract so a binding
namespace list valid on one backend is valid on all.
Delivers only the seam extension and the worker's adoption; the Python
backend, py-types renderer, and Code Mode language dispatch are later
PRs in the stack that depend on these exports.
- subagents.history computes its projections best-effort on both arms
(a hostile unit's fold rejection serves the page without the block,
matching the session-list precedent) with dual-arm coverage
- the lifecycle-witness spec probes every field of the seven-key
witness, protecting the key list itself
- list-children's own module docs catch up with the seq-gate contract,
and the design note records the rung-two later-event window as an
accepted, self-healing deviation of the corruption class
`ui-question`'s node half called `ctx.tools.register` on the host context.
`ScopedLayers.merge()` combines the global layer with the agent's exact-scope
layer, and an unscoped registration lands in the global one — so the tool
reached every agent no matter which preset composed it. `core-web`, sold as a
two-tool benchmark surface, really presented three.
Rendering a question is a host UI capability; having the tool is an agent
capability, and only a preset decides that. The node half is now empty and the
`tool-ask-user` row moved into the preset that wants it. The TUI keeps its own
row, having no presets.
The composition tests now assert the global tool layer is EMPTY, which is the
invariant that would have caught this: any tool outside a preset reaches every
agent. The browser lane's composition, seeded-history, and hermetic-skill
assertions address their registries through a composed agent for the same
reason — those services are per session now, and the host cannot resolve an
`isolate` realm by name.
The bundle split emptied `apps/cli`'s plugin dependencies, and the flat module
fallback links only that manifest's closure — so a preset row naming
`@deepseek-ai/dsh-persona` resolved to nothing and every preset mount failed,
leaving each session with an agent that had no tools, persona, or token meter.
Which packages the shipped presets compose is not implied by any bundle: the
presets live beside this manifest, so this is where their closure is declared.
The hermetic skill test now addresses the registry through the composed agent,
the only shape that can see a preset's `isolate` realm.
Moving the agent plane behind per-session presets took five rows with it that
the host still owns, and the Web surface stopped booting: `host-apiproxy`
injects `subagents`, so with the registry disabled here the entry never
activated and `dsh web` died at plugin-tree load.
The criterion is injection, not subject matter. A host row that injects a
service resolves it before any session exists, so there is no agent to key by:
`bash-env` (which `apps/cli/src/web.ts` injects to publish `DSH_WEB_URL`), the
`subagents` registry and its spawn/fork backends (a process singleton whose
cross-session queries the api-proxy serves, and whose provider names are
globally unique), and `tool-subagent-report` (a continuable setup on that
singleton, registered once per live session by a list that is not scope-aware)
all stay host-plane. What a preset chooses is which delegation TOOLS it sees.
The browser lane needs the second half: skill roots now resolve inside a preset,
a subtree the lane's include patches cannot reach, so the row's documented
environment fallback is pinned for the whole scaffold lifetime — presets mount
when a session is created, not at boot. Without it a developer's real
~/.dsh/skills enters replay requests and goldens while CI sees none.
The Agent Note was written when only the seam existed and never caught up.
Rewritten in place, per the implemented-note contract, with the four facts the
later work established:
- a preset file is an INPUT: `EntryTree.write()` persists a tree whenever the
Loader thinks the config changed, and a self-disposing plugin is enough, so
the inherited behaviour truncates a shipped preset to `[]` the first time a
session ends
- a plugin that looks itself up in the global registry breaks inside a preset,
because `register()` files into the calling context's scope — the general
rule behind the `dsh-tool-skill` fix
- an entry-local `isolate` realm is invisible to the agent's own scope too, not
only to the host, which is what makes a preset's registry that agent's own
and also why a consumer left outside the group silently contributes nothing
- switching is blank-only, and why it swaps the subtree rather than the session
`docs/architecture.md` gains an Agent Presets section: the map has to carry a
new architectural concept or it is wrong, and the root layout gains the group.
Both budget ceilings are raised rather than the content cut. `AGENTS.md` sat at
1774/1775 — one word of room, already far under the 5% headroom the standard
asks for — so no group line could be added at all; `architecture.md` was in the
same shape. Raising restores headroom instead of encoding "the map may not grow".
A third built-in preset: the standard coding agent plus the self-referential
Cordis toolset, a persona that explains the two-plane split, and a skill
teaching composition authoring. It exists so a person can ask an agent to
author another agent.
The skill ships INSIDE the preset directory rather than in the user's skill
root, and the root is derived from the preset's own `baseUrl` — the loader
evaluates `!!js` with `with (ctx)`, so a composition can locate itself. A
preset is the unit that gets copied and edited, so its documentation should
travel with it.
The skill leads with the rule that actually bites: a row publishing a service
may not sit loose in a preset, whether a row publishes one is not visible from
its name (`tool-bash` provides `bashEnv`), and a consumer left outside its
provider's isolate group resolves the host registry and then contributes
nothing — the quietest failure this design has.
Writing the test surfaced a consequence worth stating: an entry-local realm
makes the service invisible to the agent's own scope too, not just to the host.
Only rows inside that group resolve it, which is precisely what makes
`tool-skill` this agent's own rather than a shared one. The test asserts what
is actually observable from outside instead of reaching for the isolated
service.
TRUST: `cordis_mount` evaluates model-written JavaScript against the live
runtime, and a composition this agent writes becomes a preset other sessions
mount. Both the preset header and the toolset's own documentation say to treat
this as shell access. The tools stay opt-in per session — a test pins that they
are absent from every other preset.
The creation header names the preset a session STARTED with and is frozen,
which is correct — it is a creation fact. Switching is legal only while a
session is blank, and that looked like enough: no history exists yet.
It is not, because the switch's effect outlives the blank window. The user
switches, then sends the first message; every turn from there runs under the
new composition while the header still names the old one. The session is
then locked around a misrecorded preset, and resume reads the header to
rebuild it — composing one preset's tools over a history another produced,
which is exactly the replay the blank-only lock exists to prevent, reached
by another route. A picker showed `standard` for a session running
`core-web`.
A switch is now an `agent-preset/selected` event appended after the swap
commits, and `resolveSessionPreset()` (last selection, else the header) is
what every reconstruction reads: the summary, resume, the conflict guard,
and the fork introduced one layer down.
A seat in the composer tool row, left of the model select, showing the preset
THIS session runs — its own recorded one, not the deployment default, because a
resumed session runs what it was created with. `SessionSummary` carries
`agentPreset` for that, alongside `cwd` and `origin`.
The switch exists only while the conversation has not started. After the first
turn the seat becomes a plain label rather than a disabled menu: a greyed
control reads as "temporarily unavailable", when in fact the choice is gone for
good. The seat never asks in that state, and the host refuses independently
with `agent-preset-locked`, so a stale client cannot slip one past it.
A rejected switch restores the previous value and surfaces the host's message
rather than leaving the seat showing something the session is not running.
