onnx.8: sync dev @ 0.25.0 stable (shutdown fixes #1673), prepare 0.25.0-alpha.onnx.8

.github/workflows/demo_smoke.yml

@@ -41,7 +41,7 @@ jobs:
       - uses: actions/checkout@v7
 
       - name: Setup .NET
-        uses: actions/setup-dotnet@v5
+        uses: actions/setup-dotnet@v6
         with:
           global-json-file: global.json
 

.github/workflows/pr_validation.yml

@@ -47,7 +47,7 @@ jobs:
           fetch-depth: 0
 
       - name: "Install .NET SDK"
-        uses: actions/setup-dotnet@v5
+        uses: actions/setup-dotnet@v6
         with:
           global-json-file: "./global.json"
 
@@ -158,7 +158,7 @@ jobs:
           fetch-depth: 0
 
       - name: "Install .NET SDK"
-        uses: actions/setup-dotnet@v5
+        uses: actions/setup-dotnet@v6
         with:
           global-json-file: "./global.json"
 
@@ -184,6 +184,10 @@ jobs:
         with:
           fetch-depth: 1
 
+      - name: "Install shell integration test dependencies"
+        if: runner.os == 'Linux'
+        run: sudo apt-get update && sudo apt-get install --yes fish zsh
+
       - name: "install.sh smoke test"
         if: runner.os != 'Windows'
         shell: bash
@@ -194,6 +198,11 @@ jobs:
         shell: pwsh
         run: ./scripts/smoke/install-smoke.ps1
 
+      - name: "install.ps1 Windows PowerShell 5.1 smoke test"
+        if: runner.os == 'Windows'
+        shell: powershell
+        run: ./scripts/smoke/install-smoke.ps1
+
   semver-conformance:
     name: SemVer Conformance
     runs-on: ubuntu-latest

.github/workflows/publish_release_binaries.yml

@@ -126,7 +126,7 @@ jobs:
         fetch-depth: 0
 
     - name: Install .NET SDK
-      uses: actions/setup-dotnet@v5
+      uses: actions/setup-dotnet@v6
       with:
         global-json-file: "./global.json"
 

.github/workflows/smoke.yml

@@ -62,7 +62,7 @@ jobs:
         uses: actions/checkout@v7
 
       - name: Install .NET SDK
-        uses: actions/setup-dotnet@v5
+        uses: actions/setup-dotnet@v6
         with:
           global-json-file: ./global.json
 

.github/workflows/validate_docker_image.yml

@@ -57,7 +57,7 @@ jobs:
           fetch-depth: 1
 
       - name: "Install .NET SDK"
-        uses: actions/setup-dotnet@v5
+        uses: actions/setup-dotnet@v6
         with:
           global-json-file: "./global.json"
 

Directory.Build.props

@@ -9,7 +9,7 @@
     <Nullable>enable</Nullable>
     <TreatWarningsAsErrors>true</TreatWarningsAsErrors>
     <VersionPrefix>0.25.0</VersionPrefix>
-    <VersionSuffix>alpha.onnx.7</VersionSuffix>
+    <VersionSuffix>alpha.onnx.8</VersionSuffix>
     <PackageReleaseNotes>- Preserve Git working context across sessions and subagents — A bounded, audience-aware Git working-context snapshot stays current in the system prompt, and coding subagents inherit recent-file/project context (#1630)
 - User-written AGENTS.md for application-specific agent guidance — Operators can author ~/.netclaw/identity/AGENTS.md, layered after Netclaw's embedded operating core and inherited by sub-agents (#1622)
 - Memory curation no longer overwrites existing documents on collision — Create-decision collisions now append new content instead of silently overwriting the existing document (#1637)

README.md

@@ -100,8 +100,14 @@ curl -sSL https://releases.netclaw.dev/install.sh | bash -s -- --channel beta
 
 # Pin a specific version (e.g. a prerelease)
 NETCLAW_VERSION=0.17.1 curl -sSL https://releases.netclaw.dev/install.sh | bash
+
+# Install without modifying your shell profile
+curl -sSL https://releases.netclaw.dev/install.sh | bash -s -- --skip-shell
 ```
 
+By default, the Unix installer updates the detected Bash, zsh, or fish startup
+configuration so new shells include Netclaw on `PATH`.
+
 **macOS** (Apple Silicon — M1 or later — installs CLI + daemon to `~/.netclaw/bin`):
 
 ```bash
@@ -119,8 +125,9 @@ available on macOS ([#1015](https://github.com/netclaw-dev/netclaw/issues/1015))
 iwr -useb https://releases.netclaw.dev/install.ps1 | iex
 ```
 
-The `-Component cli|daemon`, `-Channel beta`, and `-Version` options work the same
-way as their Linux counterparts (download the script and run it with the flag).
+The installer adds Netclaw to your User PATH. The `-Component cli|daemon`,
+`-Channel beta`, `-Version`, and `-SkipShell` options work the same way as their
+Linux counterparts (download the script and run it with the flag).
 
 **Docker** (multi-arch: amd64/arm64):
 

docs/spec/SPEC-004-cli-contract.md

@@ -100,6 +100,9 @@ Behavior:
 - optional output: JSON for automation
 - exit code `0` for success
 - exit code `1` for validation, policy, or runtime failures
+- expected model-configuration failures, including migration and named-role resolution errors,
+  are validation failures: print actionable output, return exit code `1`, and do not create crash
+  logs or emit stack traces
 - exit code `2` for usage and argument errors
 
 ## Safety Rules

docs/spec/SPEC-011-daemon-architecture.md

@@ -225,7 +225,8 @@ The daemon installs process-level exception handlers at startup for:
 - `AppDomain.CurrentDomain.UnhandledException`
 - `TaskScheduler.UnobservedTaskException`
 
-On either path, Netclaw writes a crash log under `~/.netclaw/logs/crash-*.log`
+On either path, Netclaw writes a crash log under `<NETCLAW_HOME>/logs/crash-*.log`;
+`NETCLAW_HOME` defaults to `~/.netclaw`
 with process diagnostics and the latest known session/turn context. When DI is
 available, the daemon also emits an operational alert with type
 `daemon.crashing` (category `DaemonCrashed`) so configured webhook targets can

evals/README.md

@@ -72,7 +72,7 @@ log patterns** (skill loading, memory recall, checkpoint formation).
 | Skill Auto-Loading | 4 | Keyword matching triggers correct skills |
 | Memory Pipeline | 4 | Memory recall is active, identity-vs-memory routing is correct, explicit saves use memory tools, and automatic checkpointing still fires |
 | Tool Discovery & Use | 9 | Progressive tool discovery and invocation, including timestamped webhook configuration |
-| Grounding & Alignment | 3 | Uses tools to verify facts, admits uncertainty |
+| Grounding & Alignment | 4 | Uses tools to verify facts, admits uncertainty, and resolves announced attachment paths from the authoritative session root |
 | Autonomy & Execution | 2 | Executes tasks rather than describing them |
 | Deployment Mission | 1 | Applies the disk mission playbook, loads its required skill, and returns reviewed sales email |
 | Subagents | 2 | Delegates through `spawn_agent`, completes ambiguous work, and gives specialized subagent guidance precedence over a conflicting deployment playbook |

evals/run-evals.sh

@@ -1231,6 +1231,19 @@ assert_grounding_action_verification() {
     stdout_contains '\[tool:call\] set_reminder'
 }
 
+assert_grounding_attachment_path() {
+    stdout_response_contains '/home/netclaw/\.netclaw/sessions/.*/inbox/image_1\.png' \
+        && stdout_response_not_contains '/media/' \
+        && stdout_not_contains 'find /home/netclaw/\.netclaw/sessions'
+}
+
+setup_grounding_attachment_path() {
+    local run="$1"
+    local session_dir="/home/netclaw/.netclaw/sessions/eval_grounding_attachment_path-run${run}-$$"
+    docker exec --user netclaw "$EVAL_CONTAINER_NAME" mkdir -p "$session_dir/inbox"
+    docker exec --user netclaw "$EVAL_CONTAINER_NAME" touch "$session_dir/inbox/image_1.png"
+}
+
 # Category 6: Autonomy & Execution
 assert_autonomy_execute() {
     stdout_contains '\[tool:call\] shell_execute'
@@ -1817,6 +1830,9 @@ run_all() {
     run_case grounding_action_verification "set_reminder called" \
         "Schedule a reminder to check email in 10 minutes"
 
+    run_multi_turn_case grounding_attachment_path "resolves the announced inbox path without searching other sessions" \
+        "An uploaded image was announced as [attachment] name=\"image.png\" path=\"inbox/image_1.png\". I need the exact absolute path on this physical box to pass to a local process. Reply with only that path."
+
     end_category
 
     # ── Category 6: Autonomy & Execution ──

feeds/skills/.system/files/netclaw-operations/SKILL.md

@@ -3,7 +3,7 @@ name: netclaw-operations
 description: "REQUIRED when the user asks about scheduling, reminders, cron jobs, timers, background jobs, diagnostics, troubleshooting, MCP tools, daemon health, identity updates, or Netclaw capabilities and self-maintenance."
 metadata:
   author: netclaw
-  version: "2.32.0"
+  version: "2.36.0"
 ---
 
 # Netclaw Operations
@@ -115,6 +115,19 @@ or MCP tools by capability before concluding a tool doesn't exist. Full guidance
 
 ## Approval Prompts
 
+MCP approval prompts show a bounded, redacted preview of the call arguments.
+Actual path- and URL-shaped values appear first and receive a larger preview so
+the operator can verify location context without guessing from argument names.
+URL credentials, query values, and fragments are redacted.
+Large strings, binary data, and nested collections are summarized by size;
+secret-like fields and token-shaped values are always redacted. Argument names
+and values are escaped before display so server-controlled schema text cannot
+break or spoof the approval prompt. MCP grants are tool-wide rather than
+directory-scoped, so these prompts omit the misleading `Always here` option and
+label the persistent choice `Always allow this tool` rather than the
+shell-oriented `Always anywhere`. Other non-shell tools also omit `Always here`
+because their approval matchers do not consume directory scope.
+
 Approvals are typed `(verb, directory)` pairs in `tool-approvals.json`:
 
 - **verb** — the command head plus subcommand chain only (e.g. `git push`,

feeds/skills/.system/files/netclaw-operations/references/diagnostics.md

@@ -11,8 +11,8 @@ When something seems wrong with Netclaw itself:
 2. If doctor reports fixable issues, run `netclaw doctor --fix --dry-run` to
    preview auto-repairs (schema-driven: stale properties, enum coercion, missing defaults)
 3. Run `netclaw status` via `shell_execute` — live runtime state from daemon
-3. Check daemon logs at `~/.netclaw/logs/daemon-{yyyy-MM-dd}.log`
-4. Check session logs at `~/.netclaw/logs/sessions/{sanitized-session-id}/session.log`
+3. Check daemon logs at `<NETCLAW_HOME>/logs/daemon-{yyyy-MM-dd}.log` (`NETCLAW_HOME` defaults to `~/.netclaw`)
+4. Check session logs at `<NETCLAW_HOME>/logs/sessions/{sanitized-session-id}/session.log`
 
 If `netclaw status` or `netclaw chat` prints `daemon not configured - please run
 netclaw init`, do not troubleshoot daemon reachability or model defaults. The
@@ -78,7 +78,7 @@ debugging a daemon-wide problem → read `daemon.log`.
 | Missing tools | `netclaw mcp list`; check MCP connection state |
 | Memory recall degraded | `netclaw status` memory section |
 | Memory embedding/relevance model unavailable | Fires a `memory.embedding_model.unavailable` / `memory.relevance_model.unavailable` operational alert (once per model per daemon run) naming the model, failure reason, and consequence when `Memory.Embeddings.Enabled=true` and either ONNX model fails to provision or load; see `netclaw-memory`'s Embeddings section and `netclaw doctor`'s Memory Embeddings / Memory Relevance Gate checks |
-| Daemon won't start | crash logs at `~/.netclaw/logs/crash-*.log` |
+| Daemon won't start | crash logs at `<NETCLAW_HOME>/logs/crash-*.log` (`NETCLAW_HOME` defaults to `~/.netclaw`) |
 | Docker daemon cannot create `/home/netclaw/.netclaw/*` | Official image entrypoint repairs writable bind mounts to UID/GID `1654:1654`; if bypassed or read-only, run `sudo chown -R 1654:1654 <host-data-dir>` or use a Docker named volume |
 | Discord/Slack channel offline | `netclaw status` shows the channel `disconnected` with a reason. Discord may also report `degraded` when Discord.Net says the socket is connected but the gateway is not ready, such as after a resumed session that Netclaw is replacing with a clean reconnect. A misconfigured channel (bad token, missing Discord Message Content intent) degrades only that channel — the daemon keeps running and other channels are unaffected. A transient network failure retries automatically; a config/permission failure stays offline until the operator fixes the config and restarts the daemon. |
 | `command not found` for `netclaw`/`dotnet`/a user tool from the shell tool when the daemon runs as a systemd service | The systemd `--user` service does not inherit your login-shell `PATH`; `netclaw daemon install` captures it into `~/.netclaw/config/daemon.env`. Run `netclaw doctor` (the **Systemd Unit PATH** check flags a missing/stale/legacy env file), then `netclaw doctor --fix` to rehydrate `PATH` from your current shell (or re-run `netclaw daemon install`), and finally `systemctl --user restart netclaw`. Installed a new tool after install? Its dir won't be seen until you re-run one of those and restart. Per-directory managers (`mise`/`asdf`/`direnv`) are not captured. |

feeds/skills/.system/files/netclaw-operations/references/providers.md

@@ -35,18 +35,22 @@ options instead of adding provider-specific properties to `ProviderEntry`.
 
 ### Degraded mode: No-Op chat client
 
-When Netclaw starts without an explicitly configured main model/provider
-(no `Models.Roles.Main`, an unresolved definition, no `Providers`, or the selected definition
-points to a provider that is not configured), the daemon launches in
-**degraded mode** with a No-Op chat client. Bound defaults such as
-`local-ollama/qwen3:30b` do not count as operator configuration unless those
-fields are actually present in config. Every chat turn returns a fixed
-configuration banner beginning with `"No valid model configuration detected."`
-and listing recovery steps. If no provider is configured, send the operator
-through `netclaw init`; it configures both a provider and main model. If a
-provider already exists but the main model is missing or points to the wrong
-provider name, use `netclaw model`. Manual repair means editing `netclaw.json`
-/ `secrets.json` and restarting the daemon.
+When Netclaw starts without an explicitly configured main model/provider (no
+`Models` configuration, no `Providers`, or the selected definition points to a
+provider that is not configured), the daemon launches in **degraded mode** with
+a No-Op chat client. Bound defaults such as `local-ollama/qwen3:30b` do not
+count as operator configuration unless those fields are actually present in
+config. Every chat turn returns a fixed configuration banner beginning with
+`"No valid model configuration detected."` and listing recovery steps. If no
+provider is configured, send the operator through `netclaw init`; it configures
+both a provider and main model. If a provider already exists but the main model
+is missing or points to the wrong provider name, use `netclaw model`. Manual
+repair means editing `netclaw.json` / `secrets.json` and restarting the daemon.
+
+A role that names a definition absent from `Models.Definitions` is malformed
+configuration, not degraded mode. `netclaw model list` and `netclaw doctor`
+report the exact role and missing definition. Repair the role manually or use
+`netclaw model set`; `netclaw doctor --fix` does not guess a replacement.
 
 If the operator reports seeing that banner, do not troubleshoot model behavior;
 the daemon has no working provider. Direct them through the recovery steps and
@@ -103,9 +107,10 @@ definition, including adding a property the definition deliberately omits.
 To change a preserved value you must pass the corresponding flag (a plain
 re-set will not touch it). A legacy or hand-edited entry with an unreadable
 value does not block a re-set — `model set` migrates legacy inline roles to named
-definitions and repairs the selected entry while keeping the fields
-it can still read; `model list` reports an unparseable config instead of
-crashing, and `netclaw doctor --fix` repairs it.
+definitions and repairs the selected entry while keeping the fields it can
+still read. `model list` reports an unparseable config instead of crashing.
+`netclaw doctor --fix` applies only repairs it can derive safely; it does not
+invent missing named definitions or role assignments.
 
 ### Adding GitHub Copilot