feat: bind workspace actions to rendered state — 0.9.1

.claude-plugin/marketplace.json

@@ -7,14 +7,14 @@
   },
   "metadata": {
     "description": "Structured agent-user communication: validated forms, decision cards with recommendations, structured pushback, and progress reports \u2014 batch questions instead of asking one at a time.",
-    "version": "0.9.0"
+    "version": "0.9.1"
   },
   "plugins": [
     {
       "name": "attune-forms",
       "description": "The communication grammar for AI agents: batch independent questions into ONE validated form; offer recommendations as decision cards with rationales and per-option tradeoffs; disagree constructively via pushback cards; report multi-step progress with a blocked-item picker. Renders rich HTML where the host supports widgets and degrades cleanly to plain questions everywhere else. Powered by the attune-forms PyPI package via a bundled MCP server.",
       "source": "./plugin",
-      "version": "0.9.0",
+      "version": "0.9.1",
       "author": {
         "name": "Smart AI Memory",
         "email": "admin@smartaimemory.com"

CHANGELOG.md

@@ -6,6 +6,21 @@ follow [SemVer](https://semver.org/).
 
 ## [Unreleased]
 
+## [0.9.1] — 2026-08-29
+
+### Added
+- **Bound workspace action round trips** — optional workspace id,
+  revision, contract hash, and one-render nonce context is preserved by
+  widget and Markdown renderers. The strict collector accepts only an
+  action defined by the rendered view, requires explicit confirmation
+  where declared, and rejects stale or fabricated context without
+  authorizing or executing the action.
+- **Workspace dictionary and MCP boundaries** —
+  `workspace_from_dict` rejects unknown keys throughout the closed view
+  grammar. `elicitation_render_workspace` and
+  `elicitation_collect_workspace_action` expose the same render/collect
+  pattern as forms, with a non-mocked stdio round-trip receipt.
+
 ## [0.9.0] — 2026-08-29
 
 ### Added

README.md

@@ -25,9 +25,10 @@ claude plugin install attune-forms@attune-forms
 ```
 
 The plugin teaches the session the forms discipline (the `forms` skill)
-and serves four MCP tools — `elicitation_render_form`,
+and serves six MCP tools — `elicitation_render_form`,
 `elicitation_render_widget`, `elicitation_collect_response`,
-`elicitation_ask` — from this package via `uvx`. Decision cards,
+`elicitation_ask`, `elicitation_render_workspace`, and
+`elicitation_collect_workspace_action` — from this package via `uvx`. Decision cards,
 pushback cards, progress forms, deliberation cards, triage boards,
 confirm gates, ranking lists, and assumption reviews work out of the
 box; rich HTML renders where the host
@@ -120,6 +121,13 @@ if select_form_surface(form) == "widget":
 - **Validation** — `form_from_dict` refuses malformed definitions;
   `collect_form_response` refuses malformed answers (required fields,
   option membership) with field-level problems.
+- **Command workspaces** — `workspace_from_dict` validates a closed,
+  non-executable document grammar for intake, preview, execution, and
+  receipt views. `workspace_to_widget_html` and
+  `workspace_to_markdown` render the same action contract;
+  `collect_workspace_action` rejects unknown, stale, or mismatched
+  action envelopes. Optional revision/hash/nonce bindings are echoed
+  for the host to authorize and consume — the UI never grants authority.
 - **Intake templates** — `FormTemplate` + `FieldSlot` generate a
   workflow's intake form at ask-time from named candidate providers
   (`PROVIDERS`): tools describe what they need once, and the form

plugin/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "attune-forms",
-  "version": "0.9.0",
+  "version": "0.9.1",
   "description": "Structured agent-user communication \u2014 validated forms, decision cards, pushback, progress reports, deliberation, triage boards, confirm gates, rankings, and assumption reviews via the attune-forms MCP server.",
   "author": {
     "name": "Smart AI Memory",

plugin/skills/forms/SKILL.md

@@ -11,12 +11,20 @@ argument-hint: "<what needs deciding, e.g. 'deployment options' or 'this refacto
 > **Forms** — Gathering the independent dimensions of this decision as
 > one validated form instead of asking one question at a time.
 
-This skill drives the four `attune-forms` MCP tools:
+This skill drives the six `attune-forms` MCP tools:
 
 - `elicitation_render_widget` — form dict → interactive HTML (rich surface).
 - `elicitation_render_form` — form dict → batched plain-question payloads.
 - `elicitation_collect_response` — form + answers → validated response.
 - `elicitation_ask` — native MCP elicitation dialog, where supported.
+- `elicitation_render_workspace` — validated workspace view → widget + markdown.
+- `elicitation_collect_workspace_action` — view + action envelope → validated action.
+
+Command workspaces compose forms with typed evidence, progress, receipt,
+and action blocks. For consequential actions, render a host-generated
+revision/hash/nonce binding and validate the returned envelope before
+the host authorizes anything. A successful collection proves only that
+the action belongs to that rendered view; it never executes the action.
 
 ## When to use a multi-field form (the batching rule)
 

pyproject.toml

@@ -4,8 +4,8 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "attune-forms"
-version = "0.9.0"
-description = "Dynamic forms library: declarative FormSchema, multi-surface renderers (widget HTML, AskUserQuestion, MCP elicitation), and template-driven intake generation"
+version = "0.9.1"
+description = "Dynamic forms and command workspaces: validated multi-surface interaction documents for AI agents"
 readme = "README.md"
 requires-python = ">=3.10"
 license = {file = "LICENSE"}

src/attune_forms/__init__.py

@@ -87,14 +87,19 @@
 from attune_forms.widget import WIDGET_RESPONSE_MARKER, form_to_widget_html
 from attune_forms.workspace import (
     WorkspaceAction,
+    WorkspaceActionBinding,
     WorkspaceActionIntent,
+    WorkspaceActionResponse,
     WorkspaceBlock,
     WorkspaceBlockKind,
     WorkspaceItem,
     WorkspaceSection,
     WorkspaceTone,
+    WorkspaceValidationError,
     WorkspaceView,
     WorkspaceViewId,
+    collect_workspace_action,
+    workspace_from_dict,
     workspace_to_markdown,
     workspace_to_widget_html,
 )
@@ -139,14 +144,19 @@
     "set_keyboard_mode",
     "token",
     "WorkspaceAction",
+    "WorkspaceActionBinding",
     "WorkspaceActionIntent",
+    "WorkspaceActionResponse",
     "WorkspaceBlock",
     "WorkspaceBlockKind",
     "WorkspaceItem",
     "WorkspaceSection",
     "WorkspaceTone",
     "WorkspaceView",
     "WorkspaceViewId",
+    "WorkspaceValidationError",
+    "collect_workspace_action",
+    "workspace_from_dict",
     "workspace_to_markdown",
     "workspace_to_widget_html",
     "ASSUMPTION_RULINGS",