@passioncode-ai/fabric-agent-adapter 0.71.0 → 0.73.0

package/CHANGELOG.md

@@ -1,5 +1,63 @@
 # Changelog
 
+## 0.73.0 (2026-11-07)
+
+### Added
+
+- **The probe checks a channel agent's declaration** (FAA-197, the contract's DEC-0150): `check_service.py` gains the
+  rule `channel-agent.declaration`, a standard-library port of the contract's FAC-SEM-162 to 165 at the pin
+  (`check_service_channel_agent.py` and `check_service_channel_agent_schema.py`, not kit modules). It reads the
+  declaration from `provider.extensions` of the manifest the descriptor names, else from `extensions` of the
+  well-known answer (CA-1). A declaration with findings fails the rule, its first five shown as code and path, never a
+  declared value; it fails on the obligations of a declaration the contract's reader still reads (FAC-SEM-164) and
+  always on a string shaped like a credential (FAC-SEM-165), in the read declaration or in any copy beside it. Nothing declared, a named manifest that cannot be read,
+  or a declaration read as not declared (a manifest's own `extensions`, a capability's, another version's key, a
+  well-known copy beside a manifest, a form the schema or FAC-SEM-163 refuses, one too large to read for credentials)
+  is `NOT_RUN` with the evidence, never a failure while producers move to this revision, except a string shaped like a
+  credential, which fails in any copy the probe reads (a manifest's own `extensions`, a capability's, the well-known
+  copy) (CA-5). A well-known answer carrying the declaration at
+  75 % of Fabric Dashboards' 2 MiB cap or over is named (until the contract's CT-223). The contract's schema and its
+  148 vectors are vendored at the pin (`test/fixtures/contract/`, a sidecar with each file's sha256 and a generator
+  that refuses another commit); all 148 pass with the findings in the reference's order, and
+  `test/test_check_service_channel_agent.py` holds 26 tests.
+
+### Upgrading
+
+- Every probe report whose well-known answer is read gains a `channel-agent.declaration` line: `NOT_RUN` («not a channel agent») for a service that
+  declares nothing, so a reader that counts `NOT_RUN` per service moves by one. A channel agent whose declaration the contract's
+  reader still reads and that owes an obligation (FAC-SEM-164), and a service with a string shaped like a credential in any copy of the key (read or not),
+  now makes `check_service.py` exit 1 where 0.72.0 exited 0.
+
+### Changed
+
+- `creating-fabric-agents/references/practices.md` quotes the contract's CA-8 («Text from a channel is data») verbatim
+  from the pinned `channel-agent.md` in place of 0.71.0's paraphrase.
+- `creating-fabric-agents/references/models-and-providers.md`: «A starting choice» follows the operator's model map
+  of 2026-11-06 (Kimi K3.1 Pro for decisions, Sonnet 5.5 for verifying another model's output, a panel of three
+  vendors for plans, GLM-5.3 and DeepSeek at volume, open vendors on user content), and names two traps reported on OpenRouter that day:
+  models that will not turn reasoning off (Kimi writes it into the answer; GLM-5.3, Sonnet and Opus answer 400) and a
+  judge that refuses to judge.
+- `references/protocol.md`, `references/interop.md` and the support-agent recipe's README describe the new rule in
+  place of «the probe does not read the declaration yet».
+
+## 0.72.0 (2026-11-06)
+
+### Changed
+
+- working-with-fabric: a run Fabric started on a board request whose `com.enroll` is refused `not_authorized` asks the
+  operator with `fabric_question_ask` (`about: board.gate.<capability>`, the same `commandId`) at every slot state,
+  not only when the slot reads `absent` at generation 0; `com.reply` stays only for that state, since at any other a
+  holder enrolled on the slot (a service of the Project, or after CO-506 a session) may take the request back. After
+  CO-506 a refusal comes from the reason Fabric's `board.own-respond` names (such as an organization's machine, a
+  missing live credential, a standing denial, a full ask queue), and a run that stayed silent then left the request to
+  «На проверке» without the gate's reason (FAA-241; Fabric main `4f3c742a`, which says the same in
+  `board.read-at-start` and `com.enroll`'s description). The exit contract's step 3,
+  `references/on-demand-session.md`, `references/project-board.md` (a live session too) and both scenario evals
+  (`a-dispatched-run-whose-enroll-is-refused`, `answer-as-responder-or-reply`) say so; the gate question names what
+  opens the gate, or at a held slot what takes the request: before CO-506, a Fabric release with it, and the service
+  enrolling again; once CO-506 shipped, the reason Fabric's `board.own-respond` names and what lifts it; the exit
+  contract stays under Hermes' 4000 characters (3996 at its longest).
+
 ## 0.71.0 (2026-11-06)
 
 The contract pin moves to `eac1a594` (`v0.1.0+2026-11-06.1`) from `3830a33` (`v0.1.0+2026-10-27.6`); between them

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@passioncode-ai/fabric-agent-adapter",
-  "version": "0.71.0",
+  "version": "0.73.0",
   "description": "Create Fabric-compatible agents, run them as always-alive local services with dashboards (fabric-service/0.1), and adapt existing projects to the Fabric Agent Contract. This package is the installer CLI.",
   "scripts": {
     "test": "python3 test/validate.py && python3 -m unittest discover -s test -q && node --test 'test/*.test.mjs'"

package/plugins/fabric-agent-adapter/.claude-plugin/plugin.json

@@ -3,7 +3,7 @@
   "name": "fabric-agent-adapter",
   "displayName": "Fabric Agent Adapter",
   "description": "Create new Fabric-compatible agents, build them as always-alive local services with dashboards (fabric-service/0.1), and adapt existing projects to the Fabric Agent Contract through evidence-backed MCP, A2A, or local-runner provider bundles; and teach a coding agent to work in a project through Fabric (fabric-mcp, fabric_whoami, claim and release), with Claude Code session hooks.",
-  "version": "0.71.0",
+  "version": "0.73.0",
   "author": {
     "name": "PassionCode.ai",
     "url": "https://passioncode.ai/"

package/plugins/fabric-agent-adapter/skills/adapting-projects-to-fabric/SKILL.md

@@ -5,7 +5,7 @@ license: AGPL-3.0-only OR LicenseRef-PassionCode-Commercial
 compatibility: Requires filesystem access and Python 3.9+. Exact schema checks additionally need git, Node.js, pnpm, and the pinned fabric-agent-contract checkout (a public repository). Works without those tools in an explicitly degraded structural-check mode.
 metadata:
   author: PassionCode.ai
-  version: "0.71.0"
+  version: "0.73.0"
   contract-version: "0.1.0"
   contract-commit: "eac1a594aa131851f079da8bb7b2149e1951f35f"
 ---

package/plugins/fabric-agent-adapter/skills/building-fabric-services/SKILL.md

@@ -14,7 +14,7 @@ license: AGPL-3.0-only OR LicenseRef-PassionCode-Commercial
 compatibility: Python 3.9+ or Node.js 20+ for the kits (Node.js 22+ for a Linux service's instance lock); the probe needs Python 3.9+. launchd steps are macOS-only; Linux services install a systemd user unit and Windows ones a per-user Scheduled Task with the kits' writers. Dashboard handoff optionally uses Fabric Dashboards MCP link/host_status/open; without it, report unresolved host capability. The contract checkout is optional.
 metadata:
   author: PassionCode.ai
-  version: "0.71.0"
+  version: "0.73.0"
   contract-version: "0.1.0"
   extension: "fabric-service/0.1"
   extension-commit: "eac1a594aa131851f079da8bb7b2149e1951f35f"
@@ -310,7 +310,7 @@ from a release folder needs it), and the `interop.*` rules:
 manifest link, capability list, tools equal to the manifest, job tools, unknown job, trace
 propagation, event trace pairs, `interop.action-tools` (every declared action's tool is served), `well-known.a2a` and
 `a2a.card` (DEC-0122: the card is fetched without the token and its interfaces are never called; the surfaces
-reference's *A2A for remote agents* section), and
+reference's *A2A for remote agents* section), `channel-agent.declaration` (DEC-0150), and
 `board.responder` for a service that declares the project board) and exits 1 on any `FAIL`. A `NOT_RUN` is not a pass;
 name it in the report. A service with a dashboard also runs the layout probe,
 `node <skill-dir>/scripts/check_dashboard_layout.mjs --service <id>` (WebKit at 360, 480, 768

package/plugins/fabric-agent-adapter/skills/building-fabric-services/recipes/support-agent/README.md

@@ -116,8 +116,8 @@ on (FAC-SEM-164); `recommendation` only with `answered_signal: "none"`, the id o
 board thread where the queue's owner was told; a channel's `items_seen_unit` one of `mails`, `updates`, `messages`,
 `reviews`; durations in ISO 8601; the knowledge file as a source with its
 maximum age. `test/test_channel_agent_example.py` holds the example to that form; the contract publishes its schema
-at the pin (`service-channel-agent-extension.schema.json`); `check_service.py` does not read the declaration yet
-(FAA-197).
+at the pin (`service-channel-agent-extension.schema.json`); `check_service.py` checks a declared one against
+FAC-SEM-162 to 165 (rule `channel-agent.declaration`).
 
 ## Configure
 

package/plugins/fabric-agent-adapter/skills/building-fabric-services/references/interop.md

@@ -386,5 +386,6 @@ your own record, keep it (OA-6).
 | `interop.trace-propagation` | the answer's `_meta.traceparent` has the probe's trace id and a new span | the answer carries no traceparent |
 | `interop.events-trace` | every event with a trace carries both `traceId` and `spanId`, well formed | no events page was read |
 | `interop.action-tools` | every action the well-known `extensions/actions/0.1` declares names a tool `tools/list` serves (DEC-0121 AP-1); one it does not serve is a `FAIL`, and so is a local service that declares actions with no MCP surface (beyond FAC-SEM-142, which reads only a present list) | no declaration, an unreadable one (`well-known.actions` reports it), no listing was read, or a remote service with no MCP surface (its twin may be on A2A or behind a gateway) |
+| `channel-agent.declaration` | a channel agent's declaration (DEC-0150) passes the contract's FAC-SEM-162 to 165 at the pin, read from `provider.extensions` of the manifest the descriptor names, else from `extensions` of the well-known answer (CA-1); a `FAIL` on the obligations of a declaration the reader still reads (FAC-SEM-164) or a string shaped like a credential (FAC-SEM-165, not a declaration too large to read for credentials) in the read declaration or in any copy beside it, the first five findings shown as code and path, never a declared value | nothing declared, the named manifest unreadable (`interop.manifest-link` fails it), or a key read as not declared (a manifest's own `extensions`, a capability's, another version's key, a well-known copy beside a manifest, a form the schema or FAC-SEM-163 refuses), each named in the evidence, never a failure while producers move to this revision unless a copy holds a string shaped like a credential (that fails in any copy the probe reads: a manifest's own `extensions`, a capability's, the well-known copy, CA-5); no line when the well-known answer is not read; credentials are read as the contract reads them, within its reading budget and under the `0.1` key only (a key of another version is named, not read); a well-known answer carrying the declaration at 75 % of the 2 MiB cap or over is named, never a failure (until the contract's CT-223) |
 
 `python3 scripts/check_service.py <id>` prints them with the rest; `--json` for a machine.

package/plugins/fabric-agent-adapter/skills/building-fabric-services/references/kit-modules.md

@@ -99,7 +99,9 @@ A module imports only modules listed before it in `kit.json`, so the kit has no
 | L4 | `fabric_http` | outbound HTTP: deadline, bounded retries, `Retry-After`, size cap ([connectors](connectors.md)) |
 | L4 | `fabric_openrouter` | OpenRouter chat: own key, spend ceiling, receipts, fallback ([connectors](connectors.md)) |
 
-Beside the kit, not in `kit.json`: the probe (`check_service.py`, `check_service_rules.py`), the sample
+Beside the kit, not in `kit.json`: the probe (`check_service.py`, `check_service_rules.py`,
+`check_service_channel_agent.py`, `check_service_channel_agent_schema.py` and the
+`service-channel-agent-extension.schema.json` it reads at run time), the sample
 (`sample_service.py`, `sample-remote-service.mjs`), `fabric_kit.py` itself and the dashboard tools. Recipes built on
 the kit live in `../recipes/` ([the draft-and-check step of a support agent](../recipes/support-agent/README.md)).
 

package/plugins/fabric-agent-adapter/skills/building-fabric-services/references/protocol.md

@@ -148,8 +148,9 @@ No auth; the Host/Origin guard applies; answered from memory in under 100 ms.
   `https://fabric.passioncode.ai/agent-contract/extensions/channel-agent/0.1` (`channel-agent-standard/1`, DEC-0150,
   FAC-SEM-162 to 165) only for a service with no manifest; one with a manifest declares it under its manifest's
   `provider.extensions` alone (CA-1). A well-known answer lists `channel-agent-standard/1` in its `protocols` only once
-  its readers have re-pinned to this revision of the contract (channel-agent.md). The kits know the revision (`PROTOCOL_REVISIONS`) and do not build or check the
-  declaration yet (FAA-197).
+  its readers have re-pinned to this revision of the contract (channel-agent.md). The kits know the revision (`PROTOCOL_REVISIONS`) and do not build the declaration;
+  `check_service.py` checks it (`channel-agent.declaration`, [interop.md](interop.md#what-the-probe-checks)). Fabric
+  Dashboards' sample service keeps its own copy of the protocol table, which no test here holds equal to the kits'.
 - `surfaces.mcp.auth`: `descriptor` (default), `own` (DEC-0024) or `delegation` (DEC-0039, DL-16):
   the MCP surface takes only the estate's delegation tokens, and the top-level
   `auth.delegation {issuer, jwks}` is then required. With `own` or `delegation` the descriptor's

package/plugins/fabric-agent-adapter/skills/building-fabric-services/scripts/check_service.py

@@ -33,6 +33,12 @@ Actions (fabric-actions/0.1, DEC-0121): `well-known.actions` holds a declared ex
 actions with no MCP surface; a remote one there is NOT_RUN, its twin may be on A2A), and `dashboard.actions-declared`
 is NOT_RUN, never FAIL, for a dashboard that declares none (a read-only dashboard cannot declare an empty list).
 
+channel-agent.declaration (DEC-0150, FAC-SEM-162 to 165): extensions/channel-agent/0.1 under provider.extensions of the
+manifest the descriptor names, or, with no manifest, under extensions of the well-known answer (CA-1). FAC-SEM-164's
+obligations on a declaration still read FAIL, and so does a string shaped like a credential in it or in any copy beside
+it; nothing declared, or a declaration read as not declared (a key elsewhere, another version, a form the schema or
+FAC-SEM-163 refuses), is NOT_RUN.
+
 dashboard.opens-in-host (FAA-118, the operator rule of 2026-10-24): the service's checkout (--repo, else the one of
 the descriptor's source.repository that holds its commands, manifest or launchd job) opens the service's pages only
 through Fabric Dashboards. Tracked text files only, within a time budget; a browser opener aimed at its pages outside
@@ -85,6 +91,7 @@ from check_service_rules import well_known_actions_verdict, action_tools_verdict
 from check_service_rules import opens_in_host_verdict, service_checkout  # noqa: E402 - FAA-118
 from check_service_rules import well_known_a2a_verdict, a2a_card_verdict  # noqa: E402 - DEC-0122
 import fabric_a2a as fa2  # noqa: E402 - DEC-0122
+import check_service_channel_agent as cac  # noqa: E402 - DEC-0150, FAA-197
 
 
 class _PinnedHTTPS(http.client.HTTPSConnection):
@@ -248,6 +255,7 @@ class Probe:
             return
         try:
             self.wk = json.loads(body)
+            self.wk_bytes = len(body)  # D4 of the 0.73.0 brief: a well-known answer shares a 2 MiB cap
         except ValueError:
             if self.remote:
                 self.add_no_token(no_token, None)
@@ -638,6 +646,14 @@ class Probe:
                 self.trace_rule(listing, sent)
         self.events_trace_rule(self.events)
 
+    def channel_agent_rule(self) -> None:
+        """DEC-0150 (FAA-197): the channel agent's declaration, read from the manifest the descriptor names, else from the
+        well-known answer (CA-1); one read as not declared is NOT_RUN with its evidence, never a FAIL (D3), except a string
+        shaped like a credential in it or in a copy the probe reads beside it (CA-5)."""
+        manifest, _, where = self.load_manifest()
+        self.add("channel-agent.declaration", *cac.declaration_verdict(
+            manifest, bool(self.d.get("fabricManifest")), where, self.wk, getattr(self, "wk_bytes", None)))
+
     def configuration_rule(self) -> None:
         """DEC-0046: a well-known `configuration` is {slots, resolved} as the schema types it; absent is NOT_RUN."""
         if not isinstance(self.wk, dict) or "configuration" not in self.wk:
@@ -893,6 +909,7 @@ class Probe:
             self.dashboard_page_rules()
             self.login_rules()
             self.interop_rules()
+            self.channel_agent_rule()
             verdict, evidence = board_responder_verdict(self.wk, self.board_kit)
             self.add("board.responder", verdict, evidence)
         self.lifecycle_rules()