readyagentsdev · Sep 30, 2026 · Sep 30, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -4,6 +4,8 @@ All notable changes to ReadyAgents Core.
 
 ## Unreleased
 
+## 2.0.12 — 2026-09-30
+
 ### Added
 
 - Every approval `decision` audit line (approve and reject, classic and
@@ -12,6 +14,11 @@ All notable changes to ReadyAgents Core.
   re-baselined to the new prompt, so the re-approval's line carries the new
   digest, not the one on the original `paused` line.
 
+### Docs
+
+- `docs/cli.md` shows the pip path for the gate example
+  (`readyagents new g --from-example approval_gate`).
+
 ## 2.0.11 — 2026-09-30
 
 ### Added

diff --git a/README.md b/README.md
@@ -11,7 +11,7 @@ This repository is the free core. You keep the provider account and the bill. In
 
 ## 60-second start
 
-Requires **Python 3.11–3.14** on Linux, macOS, or Windows. Current version is **2.0.11**. Install with `pip install readyagentsdev`, or from this clone. The `2` versions the [core contract](docs/stability.md#what-the-version-number-means) only; extras carry their own maturity tiers.
+Requires **Python 3.11–3.14** on Linux, macOS, or Windows. Current version is **2.0.12**. Install with `pip install readyagentsdev`, or from this clone. The `2` versions the [core contract](docs/stability.md#what-the-version-number-means) only; extras carry their own maturity tiers.
 
 ```bash
 pip install readyagentsdev
@@ -232,7 +232,7 @@ flowchart LR
 - [Compliance evidence](docs/compliance.md) (Articles 12–14 mapping; not certification)
 - [Observability](docs/observability.md)
 - [Changelog](CHANGELOG.md)
-- [Release notes 2.0.11](RELEASE_NOTES.md)
+- [Release notes 2.0.12](RELEASE_NOTES.md)
 
 ## Install extras
 

diff --git a/RELEASE_NOTES.md b/RELEASE_NOTES.md
@@ -1,3 +1,31 @@
+# ReadyAgents Core 2.0.12
+
+**The decision line records which prompt was approved.**
+
+- Every approval `decision` audit line (approve and reject, classic and
+  enterprise gates) now carries `prompt_sha256`, the digest of the prompt the
+  decision was made on.
+- After a `reapprove_required` refusal the pause is re-baselined to the new
+  prompt, so the re-approval's `decision` line carries the new digest, not the
+  one on the original `paused` line.
+- `docs/cli.md` shows the pip path for the gate example:
+  `readyagents new g --from-example approval_gate`.
+
+## Try it
+
+```bash
+pip install readyagentsdev==2.0.12
+readyagents new /tmp/gate212 --from-example approval_gate
+readyagents run /tmp/gate212/workflow.yaml            # exit 2, paused
+# edit the gate prompt in workflow.yaml, then:
+readyagents resume <run_id> --approve gate --json    # exit 3, reason reapprove_required
+readyagents resume <run_id> --approve gate           # exit 0; decision line has the new prompt_sha256
+```
+
+Or clone https://github.com/readyagentsdev/readyagents-core and `pip install -e .`.
+
+
+
 # ReadyAgents Core 2.0.11
 
 **Approve what you saw: approval prompt digest and `reapprove_required`.**

diff --git a/docs/cli.md b/docs/cli.md
@@ -30,6 +30,8 @@ readyagents new demo --template gated
 readyagents run demo/workflow.yaml --approve gate
 readyagents new --list-examples
 readyagents new f --from-example calc_pipeline
+readyagents new g --from-example approval_gate   # pip path for examples/approval_gate.yaml
+readyagents run g/workflow.yaml
 ```
 
 Templates: `basic`, `approval`, `research` (parallel + approval), `pipeline` (default; calc/json/condition), `review` (read_file + approval), `foreach`, `agent-tools`, `gated`, `decide` (confidence-gated triage; keyless run is not calibration).

diff --git a/docs/first-ten-minutes.md b/docs/first-ten-minutes.md
@@ -1,7 +1,7 @@
 # First ten minutes
 > **Tier:** stable — core, covered by the stability contract. See [stability](stability.md).
 
-No API keys. Current version is **2.0.11** (`readyagents version`). Install from PyPI
+No API keys. Current version is **2.0.12** (`readyagents version`). Install from PyPI
 (`pip install readyagentsdev`) or from a clone (`pip install -e .`) as in the
 [README](../README.md). The wheel ships `examples/`; one road below serves both.
 The `2` versions the [core contract](stability.md#what-the-version-number-means)
@@ -144,9 +144,9 @@ Same pattern for `composed_gate` (add `--approve gate`).
 
 Full syntax: [workflows.md](workflows.md). Gate mechanics: [approvals.md](approvals.md).
 
-## Optional (2.0.11 surface)
+## Optional (2.0.12 surface)
 
-**2.0.11** includes connectors, sovereign, memory, and A2A. Keyless examples
+**2.0.12** includes connectors, sovereign, memory, and A2A. Keyless examples
 (each materialized first, same as above):
 
 ```bash

diff --git a/docs/getting-started.md b/docs/getting-started.md
@@ -9,7 +9,7 @@
 
 ## Install
 
-Current version is **2.0.11**. Preferred install is from PyPI:
+Current version is **2.0.12**. Preferred install is from PyPI:
 
 ```bash
 python -m venv .venv

diff --git a/docs/mcp-plugs.md b/docs/mcp-plugs.md
@@ -1,7 +1,7 @@
 # MCP paste catalog (client plugs)
 > **Tier:** stable — documentation only. See [stability](stability.md). Companion to [mcp.md](mcp.md).
 
-Curated **one-shot / client-side** MCP plugs for local workflow YAML. Current package **2.0.11**: `pip install "readyagentsdev[mcp]"`.
+Curated **one-shot / client-side** MCP plugs for local workflow YAML. Current package **2.0.12**: `pip install "readyagentsdev[mcp]"`.
 
 | This is | This is not |
 | --- | --- |
@@ -17,7 +17,7 @@ Prefer ReadyAgents builtins (`list_dir`, `read_file`, `write_file`, `http_get` o
 ```bash
 pip install "readyagentsdev[mcp]"
 # or: pip install -U "readyagentsdev[mcp]"
-readyagents version   # expect 2.0.11
+readyagents version   # expect 2.0.12
 ```
 
 1. Paste a block under top-level **`mcp_servers:`** in your workflow YAML.

diff --git a/docs/why-readyagents.md b/docs/why-readyagents.md
@@ -5,12 +5,12 @@ MCP connectivity is not durable orchestration.
 
 This page records two public reports about gaps between MCP wiring and durable
 workflow execution, plus one design note about how modern MCP asks a question. It
-states plainly what ReadyAgents 2.0.11 does and does not do. Each verdict is about
+states plainly what ReadyAgents 2.0.12 does and does not do. Each verdict is about
 our own software only.
 
 ## Why not LangGraph HITL / n8n MCP?
 
-If you need **hosted**, always-on MCP recovery or a managed HITL control plane, use those products' hosted surfaces. ReadyAgents **2.0.11** is local persist/resume with an explicit foreground door: an approval pause is exit **2** (not a crash) and continues with `readyagents resume … --approve` — and process death still loses the in-flight executor. The sections below give PARTIAL verdicts on each MCP surface only; they are not a claim that ReadyAgents replaces LangGraph Platform HITL or n8n's MCP Server Trigger end-to-end.
+If you need **hosted**, always-on MCP recovery or a managed HITL control plane, use those products' hosted surfaces. ReadyAgents **2.0.12** is local persist/resume with an explicit foreground door: an approval pause is exit **2** (not a crash) and continues with `readyagents resume … --approve` — and process death still loses the in-flight executor. The sections below give PARTIAL verdicts on each MCP surface only; they are not a claim that ReadyAgents replaces LangGraph Platform HITL or n8n's MCP Server Trigger end-to-end.
 
 ## LangGraph MCP surface: MCP-exposed runs remain stateless
 
@@ -24,7 +24,7 @@ Quote LangSmith docs:
 
 > The current LangGraph MCP implementation does not support sessions. Each `/mcp` request is stateless and independent.
 
-ReadyAgents 2.0.11: PARTIAL. We persist local workflow state and support resume/approval pauses. The official `io.modelcontextprotocol/tasks` extension (and the deprecated `/runs` alias) returns a durable `taskId`/`run_id` that can be polled, approved via `tasks/update`, or cancelled without holding the original request open. That is a local authenticated task handle, not a hosted recovery surface: stopping the command stops the listener and in-process executor. A record may remain on disk; nothing resumes it automatically. This comparison is about the MCP surface, not LangGraph's other surfaces.
+ReadyAgents 2.0.12: PARTIAL. We persist local workflow state and support resume/approval pauses. The official `io.modelcontextprotocol/tasks` extension (and the deprecated `/runs` alias) returns a durable `taskId`/`run_id` that can be polled, approved via `tasks/update`, or cancelled without holding the original request open. That is a local authenticated task handle, not a hosted recovery surface: stopping the command stops the listener and in-process executor. A record may remain on disk; nothing resumes it automatically. This comparison is about the MCP surface, not LangGraph's other surfaces.
 
 ## n8n MCP Server Trigger: a long synchronous call can lose the result, and HITL is unsupported
 
@@ -42,7 +42,7 @@ Quote the current MCP tools reference:
 
 > Executing workflows with multi-step forms or any kind of human-in-the-loop interactions isn't supported.
 
-ReadyAgents 2.0.11: PARTIAL. We checkpoint locally and have an approval node. A disconnected 2026 MCP client can recover a local result by polling `tasks/get` after `CreateTaskResult`, and can decide (`tasks/update`) or cooperatively cancel on the same authenticated loopback door. That does not make ReadyAgents a hosted MCP Server Trigger: the door is an explicit foreground command, and process death loses the in-flight executor. n8n's instance-level `execute_workflow` already returns an execution ID immediately; this page is only about the MCP Server Trigger surface.
+ReadyAgents 2.0.12: PARTIAL. We checkpoint locally and have an approval node. A disconnected 2026 MCP client can recover a local result by polling `tasks/get` after `CreateTaskResult`, and can decide (`tasks/update`) or cooperatively cancel on the same authenticated loopback door. That does not make ReadyAgents a hosted MCP Server Trigger: the door is an explicit foreground command, and process death loses the in-flight executor. n8n's instance-level `execute_workflow` already returns an execution ID immediately; this page is only about the MCP Server Trigger surface.
 
 ## Design note: modern MCP uses retry-shaped multi-round trips
 
@@ -56,7 +56,7 @@ Quote the Python SDK:
 
 > A resolver works on every connection. For a client on a legacy connection the SDK sends it the question directly; on a 2026-07-28 connection the SDK returns the question from the call, and the client's next attempt carries the answer.
 
-This is a design note, not a competitor gap. The ask is now a stateless retry-shaped resolver. ReadyAgents 2.0.11 surfaces a paused approval as `tasks/get` `input_required` and accepts `tasks/update` `inputResponses`, routed through the same signed/RBAC/audit path as `readyagents decide`. That is protocol-native HITL on a durable run record, not a hosted elicitation service. Process death still loses the in-flight executor. The localhost approval page remains an additional local door, not a substitute for MCP.
+This is a design note, not a competitor gap. The ask is now a stateless retry-shaped resolver. ReadyAgents 2.0.12 surfaces a paused approval as `tasks/get` `input_required` and accepts `tasks/update` `inputResponses`, routed through the same signed/RBAC/audit path as `readyagents decide`. That is protocol-native HITL on a durable run record, not a hosted elicitation service. Process death still loses the in-flight executor. The localhost approval page remains an additional local door, not a substitute for MCP.
 
 ## A2A: a projection, not a second runtime
 

diff --git a/pyproject.toml b/pyproject.toml
@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "readyagentsdev"
-version = "2.0.11"
+version = "2.0.12"
 description = "Local one-shot YAML/JSON agent workflow engine + MCP toolkit (BYOK)."
 readme = "README.md"
 license = { text = "Apache-2.0" }

diff --git a/schemas/workflow-v1.json b/schemas/workflow-v1.json
@@ -4331,5 +4331,5 @@
   ],
   "title": "ReadyAgents workflow",
   "type": "object",
-  "x-readyagents-version": "2.0.11"
+  "x-readyagents-version": "2.0.12"
 }
diff --git a/server.json b/server.json
@@ -7,13 +7,13 @@
     "url": "https://github.com/readyagentsdev/readyagents-core",
     "source": "github"
   },
-  "version": "2.0.11",
+  "version": "2.0.12",
   "packages": [
     {
       "registryType": "pypi",
       "registryBaseUrl": "https://pypi.org",
       "identifier": "readyagentsdev",
-      "version": "2.0.11",
+      "version": "2.0.12",
       "packageArguments": [
         {
           "type": "positional",

diff --git a/src/readyagents/__init__.py b/src/readyagents/__init__.py
@@ -2,7 +2,7 @@
 
 from __future__ import annotations
 
-__version__ = "2.0.11"
+__version__ = "2.0.12"
 
 from readyagents.errors import (
     ApprovalRequired,

diff --git a/src/readyagents/skills/agents_md.py b/src/readyagents/skills/agents_md.py
@@ -7,14 +7,14 @@
 HEADER = """# AGENTS.md
 
 This is a ReadyAgents Core repository: a local one-shot YAML/JSON agent workflow
-engine. Current **2.0.11** — soft path is `pip install readyagentsdev`, then
+engine. Current **2.0.12** — soft path is `pip install readyagentsdev`, then
 `readyagents new f --from-example calc_pipeline`. Always-on packs are optional
 and not required to run or validate workflows.
 
 ## Run
 
 ```sh
-# PyPI soft path (Current 2.0.11)
+# PyPI soft path (Current 2.0.12)
 pip install readyagentsdev
 readyagents new f --from-example calc_pipeline
 readyagents run f/workflow.yaml

diff --git a/tests/test_publish_mcp.py b/tests/test_publish_mcp.py
@@ -9,14 +9,14 @@
 import pytest
 
 ROOT = Path(__file__).resolve().parents[1]
-PYPI_JSON = "https://pypi.org/pypi/readyagentsdev/2.0.11/json"
+PYPI_JSON = "https://pypi.org/pypi/readyagentsdev/2.0.12/json"
 VERSION_NOT_FOUND = (
     'Error: publish failed: server returned status 400: {"title":"Bad Request",'
     '"status":400,"detail":"Failed to publish server","errors":[{"message":'
     '"registry validation failed for package 0 (readyagentsdev): PyPI package '
-    "'readyagentsdev' exists, but version '2.0.11' was not found (status: 404). "
+    "'readyagentsdev' exists, but version '2.0.12' was not found (status: 404). "
     "A newly published release can take a moment to appear on PyPI. Wait and "
-    "retry, or publish version '2.0.11' before registering it\"}]}"
+    "retry, or publish version '2.0.12' before registering it\"}]}"
 )
 OWNERSHIP_400 = (
     'Error: publish failed: server returned status 400: {"title":"Bad Request",'
@@ -61,7 +61,7 @@ def get(url: str) -> int:
 
     wait_pypi.wait_until_pypi_version_visible(
         "readyagentsdev",
-        "2.0.11",
+        "2.0.12",
         timeout=60,
         interval=10,
         get=get,
@@ -87,7 +87,7 @@ def get(url: str) -> int:
 
     wait_pypi.wait_until_pypi_version_visible(
         "readyagentsdev",
-        "2.0.11",
+        "2.0.12",
         timeout=60,
         interval=10,
         get=get,
@@ -112,7 +112,7 @@ def get(url: str) -> int:
     with pytest.raises(wait_pypi.PyPIVersionNotVisible, match="was not found") as exc:
         wait_pypi.wait_until_pypi_version_visible(
             "readyagentsdev",
-            "2.0.11",
+            "2.0.12",
             timeout=30,
             interval=10,
             get=get,
@@ -135,7 +135,7 @@ def get(url: str) -> int:
     with pytest.raises(wait_pypi.PyPIVersionNotVisible, match="last status: 500"):
         wait_pypi.wait_until_pypi_version_visible(
             "readyagentsdev",
-            "2.0.11",
+            "2.0.12",
             timeout=20,
             interval=10,
             get=get,
@@ -148,7 +148,7 @@ def test_wait_first_try_200_does_not_sleep() -> None:
     clock = _Clock()
     wait_pypi.wait_until_pypi_version_visible(
         "readyagentsdev",
-        "2.0.11",
+        "2.0.12",
         timeout=600,
         interval=15,
         get=lambda url: 200,
@@ -238,7 +238,7 @@ def get(url: str) -> int:
             "--package",
             "readyagentsdev",
             "--version",
-            "2.0.11",
+            "2.0.12",
             "--timeout",
             "60",
             "--interval",
@@ -254,7 +254,7 @@ def test_main_wait_timeout_exits_1(monkeypatch: pytest.MonkeyPatch) -> None:
     monkeypatch.setattr(wait_pypi, "http_get", lambda url: 404)
     monkeypatch.setattr(wait_pypi, "sleep", clock.sleep)
     monkeypatch.setattr(wait_pypi, "monotonic", clock.monotonic)
-    rc = wait_pypi.main(["wait", "--version", "2.0.11", "--timeout", "20", "--interval", "10"])
+    rc = wait_pypi.main(["wait", "--version", "2.0.12", "--timeout", "20", "--interval", "10"])
     assert rc == 1