@junghanacs/entwurf 0.33.0 → 0.34.0

package/AGENTS.md

@@ -31,7 +31,7 @@ Ask: does this follow capability rather than a surface name; record backend asym
 3. **One identity contract, capability domains.** Branch on rail/liveness capability, not a privileged citizen class.
 4. **Live-fact dispatch.** Never store liveness or infer send/resume before resolving the record and probing its rail.
 5. **Rejects are real.** Wrong intent, dead/drifted identity, undeliverable mailbox, ambiguity, and indeterminate probe reject without mutation or fallback.
-6. **MCP and native modules are explicit.** MCP uses only explicit `mcpServers`; native extension modules use operator-declared paths through the owning bootstrap. Neither discovers ambient configuration, scans extension directories, or retrieves modules automatically. Native modules do not emulate a foreign extension API.
+6. **MCP and native modules are explicit.** MCP uses only explicit `mcpServers`. Native extension modules enter only through the owning bootstrap, from an operator-named absolute file or an operator-named absolute directory read once at launch: top level only, exact suffix filter, deterministic order, every selected entry a regular file. Neither discovers ambient or default locations, takes module paths from settings or environment, recurses, watches, or retrieves modules automatically. Native modules do not emulate a foreign extension API.
 7. **One meta-store contract.** Record body and filename agree; active records are readable regular non-symlink V3 files; `nativeSessionId` is unique. Writers and doctor certify the active store before writing. Address-bearing reads use the addressable reader; `readStoreRecordFile`/`O_NOFOLLOW` and lstat policy both remain load-bearing. Quiesce and fresh-cut an unreadable generation; redeploy a stale unknown-backend reader. Store protocol and gates: `meta-session.ts`, `check-meta-*`, `check-fresh-cut-gate`.
 8. **GC reclaims processes, not memory.** Preserve or archive records and transcripts; do not casually delete them.
 9. **Not a second harness.** No prompt reconstruction, transcript hydration, tool-result ledger, credential mediation, or harness emulation. Backends own auth and transcript.
@@ -88,13 +88,13 @@ A new contract's test goes beside the behavior it certifies — `pi-extensions/*
 
 ```text
 implement → affected focused gates → independent review → one amendment bundle
-          → qualification once if gate/mutant/matrix changed → full floor once on frozen candidate → commit
+          → affected qualification PARTIAL once if gate/mutant/matrix changed → full floor once on frozen candidate → commit
 ```
 
-- GLG-requested branch checkpoint commits keep completed subject/proof bundles small: scoped review, affected focused checks and normal hooks/shared scanner before each checkpoint, with known RED/unmeasured boundaries recorded. A checkpoint is not frozen-candidate acceptance. The sequence above owns final candidate acceptance: required changed qualification, frozen FULL and native/LIVE/release floors remain mandatory before landing/shipment, never inherited from an interim commit.
+- GLG-requested branch checkpoint commits keep completed subject/proof bundles small: scoped review, affected focused checks and normal hooks/shared scanner before each checkpoint, with known RED/unmeasured boundaries recorded. A checkpoint is not frozen-candidate acceptance. The sequence above owns final candidate acceptance: the affected qualification PARTIAL, frozen FULL and native/LIVE/release floors remain mandatory before landing, and the prepared SHA's CI qualification body before shipment, never inherited from an interim commit.
 - Inner loop runs only affected gates; review and its corrections finish before the full floor.
 - A changed contract names a production subject and independent oracle; its focused assertion has a stable QK and exact-once mutant. Matrices change by their declared axes/cells, not appended anecdotes.
-- Qualification body is scheduled once when its surface changes; its manifest/head checks remain in the deterministic floor. A release requires the exact-SHA qualification evidence required by VERIFY, never a nearby green run.
+- Development never runs the whole qualification body. A changed gate/mutant/matrix runs its affected exact-argv groups once, after review and amendment, as PARTIAL receipts (`./run.sh check-gate-qualification --group '<argv JSON>' --receipt <absolute path outside the repo>`); the manifest head stays in the deterministic floor. The whole body runs once per cut, in the prepared SHA's CI: a prepared state awaits it and is never presumed PASS, and a release requires the exact-SHA evidence VERIFY names, never a nearby green run or a PARTIAL.
 - Freeze the candidate while the full floor runs; any worktree/index movement voids that receipt. Pre-commit is not the full floor.
 - LIVE/release floors are not lowered by a shorter inner loop. Failed gate or evidence downgrade blocks the relevant commit/release decision.
 

package/CHANGELOG.md

@@ -4,6 +4,63 @@ All notable changes to this project will be documented here. Format follows [Kee
 
 ## Unreleased
 
+## 0.34.0 - 2026-10-09
+
+### Release coordination
+
+- **One cut for #134 and #133.** Coordinated by the pi-durable citizen `20261008T113104-801621` and
+  implemented by Claude Code Opus under GLG's grant for prepare and make; npm publication is
+  coordinated with GLG separately. This section is written at prepare time: the prepared SHA's CI,
+  the preserved artifact and the tag are recorded in the GitHub release, not claimed here.
+- **Verified for this cut.** `check:full` passed on the prepare tree (719 s). The first
+  `release-gate --cut` ended RED, MUST 23/1/0: `smoke-codex-fresh-live` timed out because Codex
+  0.153.4 refused the stated Codex-side model `gpt-6-luna` for a ChatGPT account (vendor 400); that
+  run is preserved as RED. The rerun with the Codex model `gpt-5.6-terra` (Pi side unchanged) passed
+  MUST 24/0/0 and BEHAVIOR 1/0/0 in 1115 s, its own `check:full` 723 s. Before the commit,
+  `check-pack-install` (with the module-directory cell) and the `check-pi-durable-contact`/`-send`/
+  `-receive` gates against a bridge bundle emitted from the candidate passed, and an installed-package
+  fixture drove agent-config's background-bash module through a faux model: badge with the panel
+  hidden and at narrow sizes, report once, an interrupted task never re-run after `--continue`.
+  Vendor model calls there were 0 by construction; its token counts are faux accounting. Affected
+  qualification partials: 58/58 for #134 (earlier tree), 52/52 and 22/22 for #133. The whole
+  qualification body runs once, in the prepared SHA's CI.
+- **Not measured, by GLG's decision for this cut.** GLG waived the agy shipped-lane criterion: the
+  three agy doctors passed, the first native-push probe of a fresh conversation was indeterminate,
+  and agy native-push stays NOT CERTIFIED. GLG deferred the badge on GLG's own visible terminal and
+  the new post-install Claude session of the meta-bridge criterion to a later session; the host's
+  dev-linked doctor passed with five live joins, and the meta-bridge writer is unchanged since
+  0.33.0. None of these is a pass, and no 0.33.0 exception carries over.
+
+### pi-durable
+
+- **Background-task badge (#134).** The carrier runtime observes the native task graph from open to
+  close; `/tasks` now toggles only the panel. The carrier TUI draws a kind-agnostic badge of live
+  background tasks (`node.background`) as the dock's first, never-shrunk row, naming the ones not
+  running and shortening down to one ASCII cell; a one-row terminal with live background work shows
+  the badge, and with none the screen is upstream's. The overlay patch now carries one section per
+  patched carrier file (`runtime.js`, `tui.js`), and the emitter refuses any other shape.
+- **Module directory ingress (#134).** `--native-module-dir <absolute directory>` is the paired,
+  exclusive form of `--native-module`: one top-level read, `*.extension.mjs` non-dotfiles only,
+  byte-wise order, per-module guards, and named refusals for an empty directory, a non-regular
+  match, two names for one file and a cross-module extension or tool name. Hard Rule 6 now admits an
+  operator-named directory read once; ambient, default, recursive, watched and setting/environment
+  discovery stay forbidden.
+- **Deferred:** #106 no-turn retirement moved to ROADMAP as GLG's deferral, not implemented.
+  #132's unfinished Herdr plugin alignment (npm re-pin, source proof, pi-durable under Herdr) moved
+  to ROADMAP, closed not-planned, to reopen against this release's published bytes.
+
+### Verification
+
+- **Development runs no whole qualification body (#133).** `run.sh check-gate-qualification` now
+  hands its arguments to the runner's parser: `--group '<argv JSON>' --receipt <absolute path>`
+  runs one affected exact-argv group as a PARTIAL receipt, and a malformed, missing or repeated flag
+  is refused by name before any body or snapshot. Until now the wrapper dropped them, so a requested
+  PARTIAL started the whole body. Bare, it stays the whole body, which runs once per cut at the
+  prepared SHA's CI; a prepared state awaits it and is never presumed PASS. AGENTS, VERIFY and the
+  release skill retire the development rule of a local whole run when a gate, mutant or matrix
+  changes. Native/LIVE, the deterministic and package floors, the four required CI jobs and the
+  artifact acceptance are unchanged; P4 FULL and its CI repetition are not reduced here.
+
 ## 0.33.0 - 2026-10-08
 
 ### Release coordination

package/README.md

@@ -27,10 +27,10 @@ Claude and Cortex Code are ACP backends. Codex and Antigravity stay native so th
 retain their own tools and work context. Capabilities differ by rail: a shipped
 citizen is not a promise of every lifecycle operation.
 
-> **Current release: [0.32.0](https://github.com/junghan0611/entwurf/releases/tag/v0.32.0)**
-> ([`80d66f4`](https://github.com/junghan0611/entwurf/commit/80d66f4ca66f0be674b2d2bfbe20c49717c0340e)).
-> Includes the pinned Pi 1.0.4 native durable app/TUI and one explicit native-module ingress.
-> npm publication and registry byte equality are [verified](https://github.com/junghan0611/entwurf/issues/130#issuecomment-6049936030).
+> **Current release: [0.33.0](https://github.com/junghan0611/entwurf/releases/tag/v0.33.0)**
+> ([`fea9b3d`](https://github.com/junghan0611/entwurf/commit/fea9b3dc4779ac5deb34e239abc8e0b5e3e95382)).
+> Includes the pinned Pi 1.1.0 native durable app/TUI and one explicit native-module ingress;
+> `npm` `latest` is `0.33.0`. What that cut ran and did not run is in its CHANGELOG section.
 > Release history and detailed evidence belong in [CHANGELOG](./CHANGELOG.md).
 
 ## Install
@@ -88,8 +88,11 @@ models/auth, queues, SQLite, transcripts and recovery remain upstream-owned:
 [Hard Rule 17](./AGENTS.md#hard-rules), **supply the missing surface; do not become the harness**.
 
 Optional `--native-module <absolute file>` initializes trusted ESM before runtime/TUI
-imports, then installs its default native Extension after the contact. No automatic
-module discovery or classic Pi API shim; guards are not a sandbox or rollback.
+imports, then installs its default native Extension after the contact. After 0.33.0 (#134):
+`--native-module-dir <absolute directory>` loads the top-level `*.extension.mjs` files of that
+one named directory in byte-wise order instead, and the durable TUI keeps a live
+background-task badge on screen at any terminal size. No automatic module discovery or
+classic Pi API shim; guards are not a sandbox or rollback.
 Ordinary Pi's known `peerDependencies` warning reflects the separate native SDK supply.
 [Native setup](./docs/setup-clean-host.md#2b-optional-native-durable-pi-durable)
 owns flags, refusals and the extension contract.

package/VERIFY.md

@@ -86,7 +86,7 @@ that moved out of `scripts/` and into a file beside its subject would otherwise
 ### The canonical floor — two entry points
 
 - **Deterministic floor — two tiers (#70):** `pnpm check` is the everyday CORE — toolchain (lint + typecheck), the vitest lanes, and the pure-unit / behavioral-contract / source-topology gates plus the cheap static coherence checks. It prints its own total wall time; acceptance is ≤60s on the reference host `oracle` (an operator measurement, never a hard wall-clock gate on arbitrary hosts). `pnpm run check:full` is the FULL deterministic floor — the core plus the hermetic-integration and package/install tiers — and is what the frozen-candidate protocol, push CI, release-gate, and `prepublishOnly` run. Exact membership is the named `check:*` group scripts in `package.json` (the executable SSOT — this document records meaning and principles, not the command list); a gate changes tier by semantic-class decision, never by getting faster or slower. One core member assumes a build output: the beside test `pi-extensions/entwurf-control.test.ts` (reached through `check-tests-beside-behavior`) registers the COMPILED bridge and refuses by name when it is absent. `prepare` normally builds it at install; after an `--ignore-scripts` install or a bridge source change, run `pnpm run build-bridge` (in the sandbox candidate when verifying one) before the core. The FULL tier carries the qualification HEAD (`check-gate-manifests`, through `check:hermetic`); the everyday core does not, and neither tier carries the separately scheduled mutant-executing body (`check-gate-qualification`).
-- **Discriminating power of that floor:** `./run.sh check-gate-qualification` (scheduled, not ambient: standalone when a lane changes a gate/mutant/matrix, in the CI `check` job on a branch push that touched the qualification surface, and — for a cut — at the prepared SHA, where the exact-SHA release oracle requires that CI step (#132: `release-gate` no longer runs a second, local BODY; this moves only the release-mode owner — the development rule of one local run when a gate/mutant/matrix changes stays in force until #133 retires it, and the 0.33.0 cut skips it only by a one-cut exception) — no longer inside the default check chains; a semver TAG push triggers no CI at all, and the exact-SHA evidence a release quotes is whichever run at that commit CARRIES the body — the branch run when the push touched the qualification surface, else a dispatch run; a push that touched none of the qualification surface skips the body, and `workflow_dispatch -f qualify=true` plus a weekly `schedule` run it unconditionally — the exact-SHA release oracle refuses a SHA whose body step did not conclude success) re-plants committed defect mutants (`scripts/mutants/*.json`, one per closed defect class) in an isolated snapshot repo and requires each to turn its gate red **bounded and at its claimed `[QK:<claim>]` signature** — a wrong-reason red fails, a baseline-red control voids the whole group, and the runner is negative-controlled on every run (zero-match/multi-match/survived/wrong-reason/hang/control-red/impurity). This measures whether the deterministic gates still *block* what they claim to block; it is **not a new evidence level** (L0–L5 are untouched) and never substitutes for LIVE evidence. Per-cut records cite claim IDs + killed mutant IDs — "N checks passed" alone is not evidence. `check-agy-permission-matrix` complements it with the enumerated permission contract space (literal cells + stated exclusion rules, oracle independent of the SUT). The gate's HEAD is separable and IS in the floor: `check-gate-manifests` runs the runner self-test, validates the committed manifest set against the origin index, and asserts the declared lane inventory, while executing zero COMMITTED mutants (the self-test runs only its synthetic fixture mutants, in tmp fixture repos) and never snapshotting the repo — three of the five reds qualification has ever produced in CI died in that head, in under five seconds.
+- **Discriminating power of that floor:** `./run.sh check-gate-qualification` (scheduled, not ambient: the whole body runs only as the bare call — in the CI `check` job on a branch push that touched the qualification surface, and — for a cut — at the prepared SHA, where the exact-SHA release oracle requires that CI step once and the prepared state awaits it, never presumed PASS; a cut therefore batches its commits into that one prepared push (#132 took the cut's body off the local `release-gate`). Development runs no whole body (#133): a lane that changes a gate/mutant/matrix runs its affected exact-argv groups once, after review, as PARTIAL receipts — `./run.sh check-gate-qualification --group '<argv JSON>' --receipt <absolute path outside the repo>`, whose arguments reach the runner's parser unchanged and are refused there by name, before any body or snapshot, when malformed, missing or repeated. A receipt is a measurement of one group, never a qualification pass. Neither form is inside the default check chains; a semver TAG push triggers no CI at all, and the exact-SHA evidence a release quotes is whichever run at that commit CARRIES the body — the branch run when the push touched the qualification surface, else a dispatch run; a push that touched none of the qualification surface skips the body, and `workflow_dispatch -f qualify=true` plus a weekly `schedule` run it unconditionally — the exact-SHA release oracle refuses a SHA whose body step did not conclude success) re-plants committed defect mutants (`scripts/mutants/*.json`, one per closed defect class) in an isolated snapshot repo and requires each to turn its gate red **bounded and at its claimed `[QK:<claim>]` signature** — a wrong-reason red fails, a baseline-red control voids the whole group, and the runner is negative-controlled on every run (zero-match/multi-match/survived/wrong-reason/hang/control-red/impurity). This measures whether the deterministic gates still *block* what they claim to block; it is **not a new evidence level** (L0–L5 are untouched) and never substitutes for LIVE evidence. Per-cut records cite claim IDs + killed mutant IDs — "N checks passed" alone is not evidence. `check-agy-permission-matrix` complements it with the enumerated permission contract space (literal cells + stated exclusion rules, oracle independent of the SUT). The gate's HEAD is separable and IS in the floor: `check-gate-manifests` runs the runner self-test, validates the committed manifest set against the origin index, and asserts the declared lane inventory, while executing zero COMMITTED mutants (the self-test runs only its synthetic fixture mutants, in tmp fixture repos) and never snapshotting the repo — three of the five reds qualification has ever produced in CI died in that head, in under five seconds.
   - **The snapshot is source-only (#125).** It replicates the git work surface and never copies an ignored artifact from the checkout. A gate that consumes an ignored build output — the compiled `entwurf-bridge` a pi citizen registers — declares it on each of its mutants (`"build": {"argv": [...], "output": "<dir>"}`, one declaration per gate group). For that group only, the runner builds the output from the snapshot's own bytes before the control, rebuilds it after every mutation and from the restored bytes before the post-control (which must reproduce the baseline digest), never counts a build red as a kill, voids a run whose gate writes into the output, and removes the output when the group ends. The runner self-test owns this as `[QK:QUALIFY-BUILD-*]`.
 - **Live floor:** `LIVE=1 ./run.sh release-gate <scratch-project-dir> --cut` — `pnpm run check:full` + the v2-native live gates + the ACP plugin acceptance floor. It reports a **two-tier summary**:
   - **MUST tier** (release-blocking — owns the exit code; "green" applies only here): `pnpm run check:full`, `smoke-entwurf-v2-matrix-live`, `check-bridge`, `doctor-pi-provider` (#81 — `check-bridge` proves the launcher this checkout SHIPS; this proves the invocation the operator's pi provider actually EXECS, by booting it and requiring the entwurf verb set back, because `command -v` answers yes for a command that exits 127; wired as a step on 2026-08-19 after a relocated pnpm cmd-shim cost a cut sixteen LIVE steps before the same fact surfaced at `smoke-acp-bundled-mcp-live`), the resident-garden-guard zero-token half (record birth / record-keyed socket / attach-on-reopen), the `smoke-acp-*-live` ACP plugin smokes (socket-citizen / raw-turn / overlay / provider / session-reuse / carrier-augment / memory-containment / rgg / mcp / skill / bundled-mcp / v2-send), the axis wired in on 2026-07-31 that the aggregate had simply never listed (`smoke-claude-native-resume-live`), `smoke-entwurf-chain-live` — the cross-harness delivery chain (native Claude Code → pi GPT → pi ACP Sonnet → mailbox terminus) proving sender identity and replyability at every hop plus a real read receipt at the end — the integrated `smoke-mux-lifecycle-live` (its own note below), and the two OMP steps 0.16.0 wired: `smoke-omp-receive-live` (the addressed roundtrip into a live omp citizen; it reads the capability registry and decides its own outcome rather than passing by default) and `smoke-omp-fresh-live` (the clause 7 receipt the release stop now requires). (`smoke-session-id-name` is gone — #50 C3: its `--session-id`/`--name` substrate has no entwurf consumer anymore.)
@@ -203,10 +203,21 @@ Do not collapse source, package, fixture, and native-host evidence into one “g
 **Native durable contact (#129, #130 carrier).** `check-pi-durable-beside` is the deterministic
 beside-test/mutant coordinate in `check:contracts`: the adapter, the installed carrier/SDK-set
 verifier, the preflight, the owner join, the carrier's relocation grammar, resolver and
-maintainer emitter (its source reads refuse replace refs), and the bootstrap's explicit
-`--native-module` ingress (#130 P2), driven through the packaged `main`'s own seams: initialization
-before the TUI and the open, named refusals, the identity and directory guards, the contact-first
-order by reference, and the bridge spawn *spec* (not the child's inheritance). `check-pi-runtime-version` asks the
+maintainer emitter (its source reads refuse replace refs; its one patch carries exactly one
+section per patched carrier file, #134), and the bootstrap's explicit `--native-module` ingress
+(#130 P2) with its paired `--native-module-dir` (#134), driven through the packaged `main`'s own
+seams: initialization before the TUI and the open, named refusals, the identity and directory
+guards, the top-level `*.extension.mjs` filter and byte-wise order, cross-module name refusals, the
+contact-first order by reference, and the bridge spawn *spec* (not the child's inheritance). The
+carrier's task footer (#134) is observed in ONE sandboxed child that loads the committed carrier
+through its own resolver: the real carrier runtime opened and closed with no model (graph observed
+for the session, `/tasks` toggling only the panel), then continued after one native background task
+was committed into that session through the SDK's public `Conversation.commit` (the carrier's graph
+holds that node and the badge counts it, `pending` while recovered), and the carrier TUI's badge and
+`durableLayout` drawn by the published pi-tui renderer on a fake terminal across rows 1–8 × cols
+1–12, count changes and resizes, against the pinned upstream screen for the zero-task case. Its
+editor is a fixed stand-in: the real `CustomEditor`, overlays, emoji width and a consumer's own task
+are the visible-terminal and consumer subjects. `check-pi-runtime-version` asks the
 same verifier whether this checkout resolves exactly the pinned SDK set (the native set; the PATH
 host is setup's subject), and `check-pi-import-surface` asserts the carrier fence. The separate
 `check-pi-durable-contact`, `check-pi-durable-send` and `check-pi-durable-receive` are
@@ -227,7 +238,9 @@ same-sandbox neighbour as the rejected control), every reachable physical SDK ed
 version and declared name, a mixed-version tree refused by name although it still imports, a
 constructed depth-3 mismatch refused by the verifier and seen by the oracle, the public launch reaching the
 app's own model refusal before any birth — also with one installed `--native-module` initialized
-first, and a reserved extension name refused before anything opens — and the HOST kept a separate subject — a real pi 1.0.4
+first, a reserved extension name refused before anything opens, one installed `--native-module-dir`
+initializing only its top-level `*.extension.mjs` (#134) and a directory without a module refused
+before anything opens — and the HOST kept a separate subject — a real pi 1.0.4
 binary (the previous generation) beside the pinned native set is a named pi FAIL with zero Pi wiring while pi-durable still
 PASSes — not native admission. The maintainer's rebuild
 equality is `emit-pi-durable-carrier --check` against a supplied upstream clone; without one it is

package/docs/setup-clean-host.md

@@ -358,12 +358,33 @@ tree (for example an override that binds pi-durable's `pi-ai` to an older publis
   (`native-module-path-*`), a module that throws (`native-module-import-failed`), a function or
   nameless default (`native-module-export-invalid`), an extension named like one the app installs
   itself (`entwurf`, `coding-tools`, `pi-prompt`, `subagent`) or a tool named like one of Entwurf's
-  contact tools (`native-module-name-reserved`; any other duplicate is the native registry's to
-  judge), and a module whose initialization changes
+  contact tools (`native-module-name-reserved`; within one module, any other duplicate is the native
+  registry's to judge), and a module whose initialization changes
   `PI_SESSION_ID`, `PI_CODING_AGENT_DIR`, `HOME` or any `ENTWURF_*` (`native-module-env-identity-changed`)
   or the process directory (`native-module-cwd-changed`). Your module is trusted code; those
-  checks catch mistakes, they do not sandbox it. Nothing else names a module — no directory,
-  setting or profile — and a sibling opened by `entwurf_fresh_call` starts without one.
+  checks catch mistakes, they do not sandbox it. Nothing else names a module — no default
+  directory, setting or profile — and a sibling opened by `entwurf_fresh_call` starts without one.
+
+- **A module directory instead (#134, after 0.33.0).** `--native-module-dir
+  <absolute directory>` replaces `--native-module` (naming both is `native-module-ingress-exclusive`).
+  Entwurf reads that one directory once, at its top level: every `*.extension.mjs` that is not a
+  dotfile is a module; helpers such as `env.mjs` or `index.mjs`, other suffixes, subdirectories and
+  directory links are never opened. The modules initialize one by one in byte-wise name order
+  (`00-env.extension.mjs` before `10-tools.extension.mjs`), each under the same checks as above,
+  and install in that order after the contact. Also refused by name: a directory without a module
+  (`native-module-dir-empty`), a matching entry that is not a regular file
+  (`native-module-path-invalid`), two names for one file (`native-module-dir-duplicate-target`), and
+  a module repeating an earlier module's extension or tool name (`native-module-name-duplicate` —
+  the native registry would replace that extension or keep both tools). A refusal at a later module
+  leaves the earlier ones already evaluated; nothing is rolled back.
+
+- **Background-task badge (#134, after 0.33.0).** While a native background task is live, the durable
+  TUI shows `⏳ n tasks` as the first line under the transcript, whatever kind of task it is, with
+  the ones not running named (`aborting`, `completing`, `waiting`, `pending`). It never shrinks: a
+  short terminal loses the editor and footer before it, a one-row terminal shows only the badge, and
+  a narrow one shortens it down to `*`. `/tasks` only shows or hides the task panel; the badge stays.
+  With no background task the screen is unchanged. A task that has already finished leaves the
+  count — its outcome is reported by the task's own extension.
 
 - **Known warning.** Ordinary Pi may print `Host-provided extension packages must be declared in
   peerDependencies with a "*" range, not dependencies: @earendil-works/…` for Entwurf's

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@junghanacs/entwurf",
-	"version": "0.33.0",
+	"version": "0.34.0",
 	"description": "Garden-citizen dispatch substrate and meta-bridge for pi, Claude Code, Copilot CLI, OMP, Antigravity, and Codex CLI; Claude and Snowflake Cortex Code ACP backends.",
 	"license": "MIT",
 	"author": {

package/pi/pi-durable/bootstrap.mjs

@@ -50,18 +50,27 @@
 // its initialization, and its default export is a native durable Extension, handed to `openDurable`
 // by reference after the contact. Initialization runs before the TUI or the runtime is imported —
 // the runtime creates the provider runtime before it installs any extension (carrier/runtime.js
-// 140-146), so an extension object cannot carry an environment phase. Entwurf checks only what the
+// 142-148), so an extension object cannot carry an environment phase. Entwurf checks only what the
 // ingress needs: a default export object with a name, no extension name the app installs itself (a
 // same-name install REPLACES), and no tool name of the contact. Every other field is the native
 // registry's to validate. The module is trusted operator code: the identity carriers and the
 // process directory are compared across its initialization and a change is refused by name — a
-// mistake guard, not a sandbox, and nothing is rolled back. Nothing is discovered: no directory,
-// setting, environment variable or profile names a module.
+// mistake guard, not a sandbox, and nothing is rolled back. Nothing is discovered: no ambient or
+// default directory, setting, environment variable or profile names a module.
+//
+// `--native-module-dir <absolute directory>` (#134) is the paired form, exclusive with the file
+// ingress: the operator-named directory is read ONCE, top level only, and a name is a module exactly
+// when it matches `*.extension.mjs` and does not start with a dot. Every other name (helpers, tests,
+// subdirectories, directory symlinks) is ignored unopened; a matching name must resolve to a regular
+// file. The modules initialize one by one in byte-wise name order, each under the same guards; two
+// modules naming one extension or one tool are refused by name, because the native registry would
+// REPLACE the first extension or keep both tools. A refusal at module k leaves modules before it
+// evaluated: nothing is rolled back. No watch, no recursion, nothing persisted.
 //
 // Without that resolver the carrier's first pi-coding-agent internal import fails by name (an
 // unsupported entwurf-pi-dist: URL scheme), and that failure is the honest one.
-import { realpathSync, statSync } from "node:fs";
-import { isAbsolute } from "node:path";
+import { readdirSync, realpathSync, statSync } from "node:fs";
+import { isAbsolute, join } from "node:path";
 import { fileURLToPath, pathToFileURL } from "node:url";
 import { decodeOmpBootstrapPayload } from "../../mcp/entwurf-bridge/dist/pi-extensions/meta-bridge-omp.js";
 import {
@@ -83,7 +92,14 @@ export const PACKAGED_BRIDGE_ENTRY = fileURLToPath(
 
 /** Flags that take a value. Anything else but `--continue`/`-c` is an unknown argument — a bridge
  * entry included: the launch argv cannot point the contact at another bridge. */
-const VALUE_FLAGS = ["--provider", "--model", "--width", "--entwurf-bootstrap", "--native-module"];
+const VALUE_FLAGS = [
+	"--provider",
+	"--model",
+	"--width",
+	"--entwurf-bootstrap",
+	"--native-module",
+	"--native-module-dir",
+];
 /** Flags only a NEW session can honour. */
 const FRESH_ONLY_FLAGS = ["--provider", "--model", "--width", "--entwurf-bootstrap"];
 
@@ -93,6 +109,8 @@ const FRESH_ONLY_FLAGS = ["--provider", "--model", "--width", "--entwurf-bootstr
  * carrier/subagent.js:25). The native registry REPLACES an installed extension of the same name.
  */
 const RESERVED_EXTENSION_NAMES = [PI_DURABLE_EXTENSION_NAME, "coding-tools", "pi-prompt", "subagent"];
+/** A directory entry that is an operator module: `*.extension.mjs`, not a dotfile. */
+const MODULE_ENTRY = /^[^.].*\.extension\.mjs$/;
 /** The identity carriers a module's initialization must leave as it found them, every ENTWURF_* included. */
 const IDENTITY_KEYS = ["PI_SESSION_ID", "PI_CODING_AGENT_DIR", "HOME"];
 
@@ -131,6 +149,51 @@ function nativeModuleFile(value) {
 	return file;
 }
 
+/**
+ * The operator's module directory: absolute, resolving to a directory, read once at the top level. The
+ * matching entries in byte-wise name order, each resolved to a regular file and none twice. Decided
+ * before anything loads.
+ */
+function nativeModuleDirFiles(named) {
+	if (!isAbsolute(named))
+		throw refuse("native-module-path-not-absolute", `--native-module-dir takes an absolute path, got ${named}`);
+	let dir;
+	try {
+		dir = realpathSync(named);
+	} catch (error) {
+		throw Object.assign(refuse("native-module-path-invalid", `${named} does not resolve`), { cause: error });
+	}
+	if (!statSync(dir).isDirectory()) throw refuse("native-module-path-invalid", `${named} is not a directory`);
+	const names = readdirSync(dir)
+		.filter((name) => MODULE_ENTRY.test(name))
+		.sort((a, b) => Buffer.compare(Buffer.from(a), Buffer.from(b)));
+	if (names.length === 0) {
+		throw refuse("native-module-dir-empty", `${named} holds no *.extension.mjs module at its top level`);
+	}
+	const seen = new Map();
+	return names.map((name) => {
+		let file;
+		try {
+			file = realpathSync(join(dir, name));
+		} catch (error) {
+			throw Object.assign(refuse("native-module-path-invalid", `${name} in ${named} does not resolve`), {
+				cause: error,
+			});
+		}
+		if (!statSync(file).isFile()) {
+			throw refuse(
+				"native-module-path-invalid",
+				`${name} in ${named} matches *.extension.mjs but is not a regular file`,
+			);
+		}
+		if (seen.has(file)) {
+			throw refuse("native-module-dir-duplicate-target", `${seen.get(file)} and ${name} in ${named} are the same file`);
+		}
+		seen.set(file, name);
+		return file;
+	});
+}
+
 /** The module's default export, checked only for what the ingress needs and returned BY REFERENCE. */
 function nativeExtensionOf(extension, file) {
 	if (typeof extension === "function") {
@@ -187,6 +250,31 @@ async function loadNativeModule(file) {
 	return nativeExtensionOf(namespace.default, file);
 }
 
+/**
+ * Initialize the operator's modules in order. A later module that names an extension or a tool an
+ * earlier one already did is refused by name: the native registry would replace that extension, or keep
+ * both tools, without a word.
+ */
+async function loadNativeModules(files) {
+	const extensions = [];
+	const owners = new Map();
+	for (const file of files) {
+		const extension = await loadNativeModule(file);
+		const names = [
+			`extension ${extension.name}`,
+			...(Array.isArray(extension.tools) ? extension.tools : []).map((tool) => `tool ${tool?.name}`),
+		];
+		for (const name of names) {
+			if (owners.has(name)) {
+				throw refuse("native-module-name-duplicate", `${file} and ${owners.get(name)} both name ${name}`);
+			}
+		}
+		for (const name of names) owners.set(name, file);
+		extensions.push(extension);
+	}
+	return extensions;
+}
+
 /**
  * Open the durable app with the contact installed, then birth the citizen and arm its doorbell on
  * the root conversation, then admit the first input (when given) to the root once. On any failure
@@ -198,8 +286,8 @@ async function loadNativeModule(file) {
  * roots; tests also pass a connector). `open` replaces the runtime's `openDurable` for tests.
  * `bridgeEntry` defaults to this package's own compiled bridge; only a programmatic caller (the
  * checkout-only gates, which drive a private bundle) passes another one — never the launch argv.
- * `extensions` are the operator's native extensions (`--native-module`), installed after the
- * contact, by reference.
+ * `extensions` are the operator's native extensions (`--native-module` or `--native-module-dir`),
+ * installed after the contact, by reference.
  *
  * @param {{
  *   bridgeEntry?: string,
@@ -320,7 +408,19 @@ function parseArgs(argv) {
 		);
 	}
 	const moduleArg = values.get("--native-module");
-	const nativeModule = moduleArg === undefined ? {} : { nativeModule: nativeModuleFile(moduleArg) };
+	const dirArg = values.get("--native-module-dir");
+	if (moduleArg !== undefined && dirArg !== undefined) {
+		throw refuse(
+			"native-module-ingress-exclusive",
+			"--native-module and --native-module-dir are two forms of one ingress; name the file or the directory",
+		);
+	}
+	const nativeModule =
+		moduleArg !== undefined
+			? { nativeModules: [nativeModuleFile(moduleArg)] }
+			: dirArg !== undefined
+				? { nativeModules: nativeModuleDirFiles(dirArg) }
+				: {};
 	if (continueSession) return { continueSession, ...nativeModule };
 	const provider = values.get("--provider");
 	if (provider === undefined) {
@@ -368,10 +468,10 @@ function parseArgs(argv) {
  */
 export async function main(argv, { loadTui = () => import(TUI), open, contact, bridgeEntry } = {}) {
 	// Every argument refusal is decided here, before any module, the TUI or the app loads.
-	const { nativeModule, ...args } = parseArgs(argv);
+	const { nativeModules = [], ...args } = parseArgs(argv);
 	// The caller's directory, fixed before any operator code runs and handed to the app explicitly.
 	const cwd = process.cwd();
-	const extensions = nativeModule === undefined ? [] : [await loadNativeModule(nativeModule)];
+	const extensions = await loadNativeModules(nativeModules);
 	// Loaded before anything opens, as upstream main.ts imports it statically: a TUI that cannot
 	// load opens no session, mints no citizen and leaves nothing to clean up.
 	const { runDurableTui } = await loadTui();

package/pi/pi-durable/carrier/runtime.js

@@ -58,8 +58,10 @@ const context = BACKGROUND_CONTEXT;
 	                                                       
 	                                         
 	                                    
-	                                                        
+	                                                                                                  
 	                           
+	                                                                 
+	                             
  
 
                                     
@@ -213,6 +215,7 @@ export async function openDurable(options                     = {})
 			conversations: summaries,
 			models: models(),
 			notices: [],
+			tasksPanel: true,
 		};
 		const listeners = new Set            ();
 		let notifying = false;
@@ -330,17 +333,8 @@ export async function openDurable(options                     = {})
 					const thinking                     = agentOf(state.conversation).thinkingLevel ?? "off";
 					await current.configure({ model: ref, thinkingLevel: clampThinkingLevel(model, thinking) }, context);
 				}),
-			toggleTasks: () =>
-				command(async () => {
-					if (tasks !== undefined) {
-						closeTasks();
-						update({ tasks: undefined });
-						return;
-					}
-					const graph = await opened.taskGraph(context);
-					tasks = graph;
-					unsubscribeTasks = graph.subscribe((value) => update({ tasks: value }));
-				}),
+			// Only the panel: the graph stays observed, so live background work is never out of view.
+			toggleTasks: () => command(async () => update({ tasksPanel: !state.tasksPanel })),
 			switchConversation: (id) =>
 				command(async () => {
 					const next = await opened.conversation(id, context);
@@ -360,8 +354,11 @@ export async function openDurable(options                     = {})
 			notice("warning", `Saved model is unavailable: ${saved.provider}/${saved.modelId}`);
 		}
 		if (initial?.fallbackMessage !== undefined) notice("info", initial.fallbackMessage);
-		// The task panel starts open; /tasks hides it.
-		await controller.toggleTasks();
+		// The task graph is observed until close; its panel starts open and /tasks hides it.
+		const graph = await opened.taskGraph(context);
+		tasks = graph;
+		update({ tasks: graph.value });
+		unsubscribeTasks = graph.subscribe((value) => update({ tasks: value }));
 		// Recovered work from an interrupted turn continues now.
 		harness.resume();
 

package/pi/pi-durable/carrier/tui.js

@@ -23,12 +23,14 @@ import {
 	SelectList,
 	                     
 	Spacer,
+	                
 	setCapabilityOverrides,
 	setKeybindings,
 	Text,
 	TruncatedText,
 	TuiAltScreen,
 	VStack,
+	visibleWidth,
 } from "@earendil-works/pi-tui";
 import { getAgentDir } from "entwurf-pi-dist:config.js";
 import { KeybindingsManager } from "entwurf-pi-dist:core/keybindings.js";
@@ -150,10 +152,75 @@ class CompactionComponent extends Box {
 	                      
  
 
+/**
+ * The live background-task badge: the dock's first row, never shrunk, so live background work stays on
+ * screen however small the terminal gets. It counts the graph's background tasks whatever their kind and
+ * names the ones that are not running; with none it renders nothing and the dock is upstream's.
+ */
+export class TaskBadge                      {
+	#count = 0;
+	#detail = "";
+
+	get count()         {
+		return this.#count;
+	}
+
+	set(graph                       )       {
+		const nodes = Object.values(graph?.tasks ?? {}).filter((node) => node.background);
+		const tally = (state        , match                                  )           => {
+			const n = nodes.filter(match).length;
+			return n === 0 ? [] : [`${n} ${state}`];
+		};
+		this.#count = nodes.length;
+		this.#detail = [
+			...tally("aborting", (node) => node.abortRequested),
+			...tally("completing", (node) => !node.abortRequested && node.state.status === "completing"),
+			...tally("waiting", (node) => !node.abortRequested && node.state.status === "waiting"),
+			...tally("pending", (node) => !node.abortRequested && node.state.status === "pending"),
+		].join(", ");
+	}
+
+	/** The widest form that fits, padded when there is room; narrower terminals drop detail, then words. */
+	render(width        )           {
+		const n = this.#count;
+		if (n === 0) return [];
+		const tasks = `⏳ ${n} ${n === 1 ? "task" : "tasks"}`;
+		const detailed = this.#detail ? [`${tasks} (${this.#detail})`] : [];
+		const forms = [...detailed, tasks, `⏳ ${n}`, `⏳${n}`, `*${n}`, "*+", "*"];
+		for (const form of forms) {
+			if (visibleWidth(form) < width) return [theme.fg("accent", ` ${form}`)];
+			if (visibleWidth(form) === width) return [theme.fg("accent", form)];
+		}
+		return [theme.fg("accent", "*")];
+	}
+
+	invalidate()       {}
+}
+
+/**
+ * The screen: the transcript above the dock. The badge is the dock's first row and never shrinks, so the
+ * renderer, which clips from the bottom, drops the editor and the footer before it; a one-row terminal
+ * with live background work shows the badge instead of the transcript.
+ */
+export function durableLayout(transcript           , badge           , dock                       )         {
+	return new VStack([
+		{
+			component: transcript,
+			basis: 0,
+			grow: 1,
+			shrink: 1,
+			minSize: 1,
+			visible: (viewport) => badge.count === 0 || viewport.height > 1,
+		},
+		{ component: new VStack([{ component: badge, shrink: 0 }, ...dock]), basis: "auto", grow: 0, shrink: 1, minSize: 1 },
+	]);
+}
+
 class DurableTui {
 	static          #renderers                                = createAllToolRenderers();
 	         #ui              ;
 	         #chat = new Container();
+	         #taskBadge = new TaskBadge();
 	         #tasks = new Container();
 	         #queue = new Container();
 	         #notices = new Container();
@@ -216,15 +283,9 @@ class DurableTui {
 		content.addChild(new Spacer(1));
 		const transcript = new ScrollView(content, { follow: "end", primary: true, overscroll: "chain" });
 		this.#transcript = transcript;
-		const dock = new VStack([
-			{ component: this.#tasks, shrink: 1, minSize: 0 },
-			{ component: this.#queue, shrink: 1, minSize: 0 },
-			{ component: this.#notices, shrink: 1, minSize: 0 },
-			{ component: this.#editorContainer, shrink: 1, minSize: 3 },
-			{ component: this.#footer, shrink: 1, minSize: 0 },
-		]);
 		for (const component of [
 			this.#chat,
+			this.#taskBadge,
 			this.#tasks,
 			this.#queue,
 			this.#notices,
@@ -234,9 +295,12 @@ class DurableTui {
 			this.#ui.addChild(component);
 		}
 		this.#ui.setLayoutRoot(
-			new VStack([
-				{ component: transcript, basis: 0, grow: 1, shrink: 1, minSize: 1 },
-				{ component: dock, basis: "auto", grow: 0, shrink: 1, minSize: 1 },
+			durableLayout(transcript, this.#taskBadge, [
+				{ component: this.#tasks, shrink: 1, minSize: 0 },
+				{ component: this.#queue, shrink: 1, minSize: 0 },
+				{ component: this.#notices, shrink: 1, minSize: 0 },
+				{ component: this.#editorContainer, shrink: 1, minSize: 3 },
+				{ component: this.#footer, shrink: 1, minSize: 0 },
 			]),
 		);
 		this.#ui.setFocus(this.#editor);
@@ -299,7 +363,8 @@ class DurableTui {
 				);
 			}
 		}
-		this.#syncTasks(view.tasks);
+		this.#taskBadge.set(view.tasks);
+		this.#syncTasks(view.tasksPanel ? view.tasks : undefined);
 		this.#syncQueue((view.conversation.docs["pi.inbox"] ?? { items: [] })              );
 		this.#syncNotices(view);
 		this.#editor.borderColor = theme.getThinkingBorderColor(agentOf(view.conversation).thinkingLevel ?? "off");