Replace end-of-session mcpscripts validation with bash to prevent ~5min MCP inactivity timeouts

.github/workflows/daily-cli-performance.md

@@ -87,25 +87,27 @@ This workflow imports `shared/go-make.md` which provides:
 - **mcpscripts-go** - Execute Go commands (e.g., args: "test ./...", "build ./cmd/gh-aw")
 - **mcpscripts-make** - Execute Make targets (e.g., args: "build", "test-unit", "bench")
 
-**IMPORTANT**: Always use these mcp-script tools for Go and Make commands instead of running them directly via bash.
+**IMPORTANT**: Use **bash** for all benchmark and validation commands in this workflow. MCP connections time out after ~5 minutes of inactivity; the analysis phases in this workflow may exceed that threshold, causing end-of-phase MCP calls to fail with `MCP error -32003: context canceled`. Always prefer `make <target>` or `go <args>` in bash over `mcpscripts-*` tools in this workflow.
 
 ## Phase 1: Run Performance Benchmarks
 
 ### 1.1 Run Compilation Benchmarks
 
-Run the benchmark suite and capture results using the **mcpscripts-make** tool:
+Run the benchmark suite and capture results using **bash** (not mcpscripts — the MCP connection may time out during the analysis phases that follow):
 
 **Step 1**: Create directory for results
 
 ```bash
 mkdir -p /tmp/gh-aw/benchmarks
 ```
 
-**Step 2**: Run benchmarks using mcpscripts-make
+**Step 2**: Run benchmarks using bash
 
-Use the **mcpscripts-make** tool with args: "bench-performance" to run the critical performance benchmark suite.
+```bash
+make bench-performance
+```
 
-This will execute `make bench-performance` which runs targeted performance benchmarks and saves results to `bench_performance.txt`.
+This runs targeted performance benchmarks and saves results to `bench_performance.txt`.
 
 The targeted benchmarks include:
 - **Workflow compilation**: CompileSimpleWorkflow, CompileComplexWorkflow, CompileMCPWorkflow, CompileMemoryUsage

.github/workflows/go-logger.md

@@ -65,13 +65,16 @@ features:
 
 You are an AI agent that improves Go code by adding debug logging statements to help with troubleshooting and development.
 
-## Available Safe-Input Tools
+## Validation Commands
 
-This workflow imports `shared/go-make.md` which provides:
-- **mcpscripts-go** - Execute Go commands (e.g., args: "test ./...", "build ./cmd/gh-aw")
-- **mcpscripts-make** - Execute Make targets (e.g., args: "build", "test-unit", "lint", "recompile")
+Use **bash** for all build, test, and validation commands in this workflow. Do **not** use `mcpscripts-make` or `mcpscripts-go` for validation — MCP connections can time out after ~5 minutes of inactivity during long file-exploration phases, causing end-of-session validation to fail with `MCP error -32003: context canceled`.
 
-Use these tools for consistent execution instead of running commands directly via bash.
+```bash
+make build          # Build the project
+make test-unit      # Run unit tests
+make recompile      # Recompile workflows
+DEBUG=* ./gh-aw compile dev  # Test workflow compilation with debug logging
+```
 
 ## Efficiency First: Check Cache
 
@@ -219,24 +222,33 @@ For each file:
    - Don't over-log - focus on the most useful information
    - Ensure messages are meaningful and helpful for debugging
 
-### 5. Validate Changes
+### 5. Early Validation (After First File)
+
+After editing your **first file**, run a quick build to catch compilation errors early — before spending time on more files:
+
+```bash
+make build
+```
+
+This surfaces syntax errors or import issues immediately, saving time if the first edit has a problem.
 
-After adding logging to the selected files, **validate your changes** before creating a PR:
+### 6. Complete Validation (After All Files)
+
+After adding logging to **all selected files**, validate your changes before creating a PR:
 
 1. **Build the project to ensure no compilation errors:**
-   Use the mcpscripts-make tool with args: "build"
-
-   This will compile the Go code and catch any syntax errors or import issues.
+   ```bash
+   make build
+   ```
+   This compiles the Go code and catches any syntax errors or import issues.
 
 2. **Run unit tests to ensure nothing broke:**
-   Use the mcpscripts-make tool with args: "test-unit"
-
+   ```bash
+   make test-unit
+   ```
    This validates that your changes don't break existing functionality.
 
 3. **Test the workflow compilation with debug logging enabled:**
-   Use the mcpscripts-go tool with args: "run ./cmd/gh-aw compile dev"
-
-   Or you can run it directly with bash if needed:
    ```bash
    DEBUG=* ./gh-aw compile dev
    ```
@@ -246,9 +258,11 @@ After adding logging to the selected files, **validate your changes** before cre
    - Debug logging from your changes appears in the output
 
 4. **If needed, recompile workflows:**
-   Use the mcpscripts-make tool with args: "recompile"
+   ```bash
+   make recompile
+   ```
 
-### 6. Create Pull Request
+### 7. Create Pull Request
 
 After validating your changes:
 
@@ -322,6 +336,7 @@ Before creating the PR, verify:
 - [ ] No duplicate logging with existing logs
 - [ ] Import statements are properly formatted
 - [ ] Changes validated with `make build` (no compilation errors)
+- [ ] Changes validated with `make test-unit` (tests pass)
 - [ ] Workflow compilation tested with `DEBUG=* ./gh-aw compile dev`
 
 ## Important Notes

.github/workflows/shared/go-make.md

@@ -26,9 +26,19 @@ mcp-scripts:
       make $INPUT_ARGS
 ---
 
-**IMPORTANT**: Always use the `mcpscripts-go` and `mcpscripts-make` tools for Go and Make commands instead of running them directly via bash. These mcp-script tools provide consistent execution and proper logging.
+**IMPORTANT — bash vs. MCP for validation:**
 
-**Correct**:
+- **Use `mcpscripts-go` / `mcpscripts-make` early in a session** (within the first few minutes) for consistent, logged execution of Go and Make commands.
+- **Use direct bash commands for end-of-session validation** (i.e., after any file-exploration or analysis phase that may have taken more than ~5 minutes). MCP connections time out after ~5 minutes of inactivity; calling `mcpscripts-*` at the end of a long session fails with `MCP error -32003: context canceled`.
+
+**Validation via bash (preferred at end of session or after long exploration):**
+```bash
+make build
+make test-unit
+make recompile
+```
+
+**MCP tools (use early in session, before long exploration phases):**
 ```
 Use the mcpscripts-go tool with args: "test ./..."
 Use the mcpscripts-make tool with args: "build"
@@ -39,8 +49,7 @@ Use the mcpscripts-make tool with args: "test-unit"
 **Incorrect**:
 ```
 Use the go mcp-script tool with args: "test ./..."  ❌ (Wrong tool name - use mcpscripts-go)
-Run: go test ./...  ❌ (Use mcpscripts-go instead)
-Execute bash: make build  ❌ (Use mcpscripts-make instead)
+Use the mcpscripts-make tool with args: "build"     ❌ (after a long exploration phase — use bash instead)
 ```
 
 <!--

AGENTS.md

@@ -1109,6 +1109,42 @@ When writing workflows that use `cache-memory` to persist data across runs, be a
 
 ## Key Features
 
+### MCP Connection Inactivity Timeout — Use Bash for End-of-Session Validation
+
+**⚠️ CRITICAL WORKFLOW AUTHORING RULE: Use bash, not MCP tools, for build/test validation at the end of a session.**
+
+MCP connections (HTTP/WebSocket transports) time out after approximately **5 minutes of inactivity**. Workflows with long file-exploration or analysis phases routinely exceed this threshold. When the agent finally attempts an end-of-session validation call via an MCP tool (e.g., `mcpscripts-make build`, `mcpscripts-go test ./...`), the MCP transport has been torn down, resulting in:
+
+```
+MCP error -32003: context canceled
+```
+
+This causes the entire workflow session to fail at the last step, wasting all preceding work.
+
+**Rule**: Always use direct `bash` commands for build, test, and validation steps — especially for any step that runs *after* a file-exploration or analysis phase:
+
+```bash
+# ✅ CORRECT — use bash for validation
+make build
+make test-unit
+make recompile
+make agent-finish
+go test ./...
+```
+
+```text
+# ❌ INCORRECT — MCP tools fail after long inactivity
+Use the mcpscripts-make tool with args: "build"     ← may fail with context canceled
+Use the mcpscripts-go tool with args: "test ./..."  ← may fail with context canceled
+```
+
+**Additional rule**: Add an **intermediate validation checkpoint** using bash after the first major code edit (e.g., `make build`), not just at the very end of the session. This surfaces compile errors early, before the agent spends more context on subsequent edits.
+
+**When `mcpscripts-*` tools are safe to use:**
+- Early in a session, before any long exploration phase
+- For short-lived workflows with no significant idle time between tool calls
+- For non-validation operations that are inherently early in the session (e.g., fetching PRs with `mcpscripts-gh`)
+
 ### MCP Server Management
 ```bash
 gh aw mcp list                    # List workflows with MCP servers

pkg/actionpins/data/action_pins.json

@@ -168,10 +168,10 @@
       "version": "v2.2.0",
       "sha": "0c5077e51419868618aeaa5fe8019c62421857d6"
     },
-    "ruby/setup-ruby@v1.305.0": {
+    "ruby/setup-ruby@v1.306.0": {
       "repo": "ruby/setup-ruby",
-      "version": "v1.305.0",
-      "sha": "0cb964fd540e0a24c900370abf38a33466142735"
+      "version": "v1.306.0",
+      "sha": "c4e5b1316158f92e3d49443a9d58b31d25ac0f8f"
     },
     "super-linter/super-linter@v8.6.0": {
       "repo": "super-linter/super-linter",

pkg/workflow/data/action_pins.json

@@ -168,10 +168,10 @@
       "version": "v2.2.0",
       "sha": "0c5077e51419868618aeaa5fe8019c62421857d6"
     },
-    "ruby/setup-ruby@v1.305.0": {
+    "ruby/setup-ruby@v1.306.0": {
       "repo": "ruby/setup-ruby",
-      "version": "v1.305.0",
-      "sha": "0cb964fd540e0a24c900370abf38a33466142735"
+      "version": "v1.306.0",
+      "sha": "c4e5b1316158f92e3d49443a9d58b31d25ac0f8f"
     },
     "super-linter/super-linter@v8.6.0": {
       "repo": "super-linter/super-linter",