Prep for v2.24.1 release (#38)

* Fix release workflow: use setup-codeql-environment action instead of codeql-action/init

Co-authored-by: data-douser <70299490+data-douser@users.noreply.github.com>

* Add "fetch-tags" to release.yml workflow checkout

* Upgrade codeql and repo version to 2.24.1

* Upgrade NodeJS dependencies to latest

* Enhance MCP prompts with LSP tool guidance

This commit:

- updates 'server/src/prompts/*.prompt.md` files to provide better
  guidance to LLMs in relation to iterative and/or LSP-based tools;
- improves unit tests of MCP ^ "workflow prompts";
- adds an example workshop, focused on using LSP-based MCP server
  tools, under the existing create-codeql-query-development-workshop
  agent skill;
- updates .github/ instructions and prompts with lessons learned
  from tool validation via example workshop development.

* improve release.yml workflow_dispatch compatibility

* Improve prettier fix markdownlint required workflow

* Fix tests flagged in PR review comments

* Fix `release.yml` workflow for publishing CodeQL packs and npm package for `codeql-development-mcp-server` (#40)

* Initial plan

* Enable auto-creation of tags in release workflow for workflow_dispatch

Co-authored-by: data-douser <70299490+data-douser@users.noreply.github.com>

* More fixes for release workflow

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>

* Add `advanced-security/ql-mcp-*-tools-src` query pack archives as uploaded release artifacts (#41)

* Initial plan

* Enable auto-creation of tags in release workflow for workflow_dispatch

Co-authored-by: data-douser <70299490+data-douser@users.noreply.github.com>

* More fixes for release workflow

* Add CodeQL query packs as uploaded release artifacts

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>

* Add version update script & fix packs for release (#42)

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>

Author: data-douser

.codeql-version

@@ -1 +1 @@
-v2.24.0
+v2.24.1

…hub/instructions/prompts.instructions.md …tructions/github_prompts.instructions.md.github/instructions/prompts.instructions.md renamed to .github/instructions/github_prompts.instructions.md .github/instructions/prompts.instructions.md renamed to .github/instructions/github_prompts.instructions.md

@@ -1,9 +1,9 @@
 ---
-applyTo: '**/*.prompt.md'
+applyTo: '.github/prompts/*.prompt.md'
 description: 'Instructions for managing .prompt.md files throughout the advanced-security/codeql-development-mcp-server repository.'
 ---
 
-# Instructions for managing `*.prompt.md` files
+# Instructions for managing `.github/prompts/*.prompt.md` files
 
 ## PURPOSE
 
@@ -27,13 +27,11 @@ That higher-level component should link (i.e. point) to at least one `.github/pr
 - ALWAYS follow best practices for writing markdown files, including proper use of headings, lists, links, and code blocks. This explicitly includes inserting a newline before and after code blocks, lists, and headings.
 - ALWAYS check formatting with `npm run lint && npm run format:check` from the repo root directory to ensure consistent formatting after making changes.
 - ALWAYS fix linting and formatting errors by running `npm run lint:fix && npm run format` from the repo root directory before committing changes.
-- ALWAYS start each `*.prompt.md` file with the following YAML block of frontmatter:
-
-  ```yaml
-  ---
-  mode: agent
-  ---
-  ```
+- ALWAYS start each `*.prompt.md` file with a YAML front-matter block containing, at minimum, values for fields such as:
+  - `agent` -> pointing to the name of the agent this prompt is intended for (e.g. `agent: ql-mcp-tool-tester`)
+  - `name` -> a unique name for the prompt (e.g. `name: validate-ql-mcp-tools-via-workshop`)
+  - `description` -> a concise description of the prompt's purpose and functionality
+  - `argument-hint` -> a brief hint about the expected arguments for the prompt
 
 ## PREFERENCES
 

.github/instructions/server_src_prompts_md.instructions.md

@@ -0,0 +1,44 @@
+---
+applyTo: 'server/src/prompts/*.prompt.md'
+description: 'Instructions for MCP-server-hosted workflow prompt files.'
+---
+
+# Copilot Instructions for `server/src/prompts/*.prompt.md` prompt files
+
+## PURPOSE
+
+This file contains instructions for working with workflow prompt files in the `server/src/prompts/` directory. These prompts are registered by the MCP server (via `workflow-prompts.ts`) and exposed as slash commands in VS Code Copilot Chat.
+
+## REQUIREMENTS
+
+- ALWAYS start each prompt file with the following YAML frontmatter block:
+
+  ```yaml
+  ---
+  agent: agent
+  ---
+  ```
+
+  Note: VS Code has deprecated `mode: agent` in favor of `agent: agent`.
+
+- ALWAYS ensure each prompt has a corresponding registration in `server/src/prompts/workflow-prompts.ts`, including a Zod parameter schema and a `server.prompt()` call.
+- ALWAYS ensure the prompt name in `WORKFLOW_PROMPT_NAMES` matches the registration.
+- ALWAYS write prompts that work in **any environment** where `codeql` and the MCP server tools are available, including terminal-only environments without an IDE.
+- ALWAYS use explicit, numeric tool parameters (e.g., `file_path`, `line`, `character`) instead of IDE-centric language like "position the cursor" or "click on".
+- ALWAYS document when MCP tools use **0-based** positions (all `codeql_lsp_*` tools) versus **1-based** positions (`find_predicate_position`, `find_class_position`, `read_file`).
+- ALWAYS note the `workspace_uri` requirement for LSP tools: it must be a **plain directory path** to the pack root containing `codeql-pack.yml`, not a `file://` URI.
+- **ALWAYS run `npm run tidy` from the repo root directory to apply (markdown) linting for all prompt files.**
+
+## PREFERENCES
+
+- PREFER referencing other prompts by their registered MCP name (e.g., `codeql://prompts/ql_lsp_iterative_development`) to enable cross-prompt navigation.
+- PREFER including a "Worked Example" section showing concrete tool invocations with realistic parameter values.
+- PREFER a validation tools comparison table when multiple tools can validate queries at different fidelity levels (e.g., `validate_codeql_query` vs `codeql_lsp_diagnostics` vs `codeql_query_compile`).
+- PREFER documenting tool limitations discovered through actual usage (e.g., `codeql_lsp_diagnostics` cannot resolve imports; `find_class_position` finds `class` only, not `module`).
+
+## CONSTRAINTS
+
+- NEVER leave any trailing whitespace on any line.
+- NEVER use four backticks to nest one code block inside another.
+- NEVER assume the calling LLM has access to an IDE, cursor, or editor UI.
+- NEVER reference `file://` URIs for the `workspace_uri` parameter — use plain directory paths.

.github/prompts/validate-ql-mcp-server-tools-via-workshop.prompt.md

@@ -3,7 +3,7 @@ agent: mcp-enabled-ql-workshop-developer
 name: validate-ql-mcp-server-tools-via-workshop
 description: 'A prompt for validating the real-world functionality of the CodeQL Development MCP Server tools by creating a CodeQL query development workshop from scratch using an existing, production-grade CodeQL query as the workshop "solution".'
 argument-hint: 'Provide the absolute or relative path to a local ".ql" or ".qlref" file associated with a production-grade CodeQL query to be used as the "solution" for the last stage of the to-be-created workshop.'
-model: Claude Opus 4.5 (copilot)
+model: Claude Opus 4.6 (copilot)
 ---
 
 # `validate-ql-mcp-server-tools-via-workshop` Prompt

.github/skills/create-codeql-query-development-workshop/examples/codeql-sap-js-ui5-xss/README.md

@@ -0,0 +1,201 @@
+# JavaScript XSS Taint-Tracking Workshop
+
+A CodeQL query development workshop that teaches you to find client-side cross-site scripting (XSS) vulnerabilities using taint-tracking analysis. Inspired by a production SAP UI5 XSS query pattern.
+
+## Background
+
+This workshop is modeled after a real-world production query (`UI5Xss.ql`) that detects XSS vulnerabilities in SAP UI5 applications. The production query uses the `DataFlow::ConfigSig` pattern with:
+
+- **Sources**: `RemoteFlowSource` and DOM-based XSS sources
+- **Sinks**: HTML injection sinks (`innerHTML`, `document.write`, etc.)
+- **Barriers**: Sanitizer functions (`encodeHTML`, `encodeJS`, etc.)
+- **Path-problem**: Full taint path visualization
+
+Since the production query depends on custom SAP libraries, this workshop uses **only** `codeql/javascript-all` (version `2.6.19`) to teach the same core concepts with standard JavaScript APIs.
+
+## Prerequisites
+
+- CodeQL CLI v2.23.9 or later
+- `codeql/javascript-all` pack (installed automatically via `codeql pack install`)
+
+## Setup
+
+```bash
+# Navigate to the workshop directory
+cd .github/skills/create-codeql-query-development-workshop/examples/codeql-sap-js-ui5-xss
+
+# Install dependencies for all packs
+codeql pack install solutions
+codeql pack install solutions-tests
+codeql pack install exercises
+codeql pack install exercises-tests
+
+# Build test databases (optional — tests do this automatically)
+chmod +x build-databases.sh
+./build-databases.sh
+```
+
+## Workshop Structure
+
+```text
+exercises/          — Your workspace (incomplete queries to complete)
+exercises-tests/    — Tests that validate your solutions
+solutions/          — Reference solutions (don't peek!)
+solutions-tests/    — Tests for the solutions
+tests-common/       — Shared test source code
+```
+
+## Test Code Overview
+
+The test file (`tests-common/test.js`) contains carefully crafted cases:
+
+| Case       | Function                  | Source                          | Sink                                   | Expected                 |
+| ---------- | ------------------------- | ------------------------------- | -------------------------------------- | ------------------------ |
+| POSITIVE 1 | `positiveDirectInnerHTML` | `document.location.search`      | `innerHTML`                            | Detected                 |
+| POSITIVE 2 | `positiveDocumentWrite`   | `window.location.hash`          | `document.write()`                     | Detected                 |
+| POSITIVE 3 | `positiveOuterHTML`       | URL param via `URLSearchParams` | `outerHTML`                            | Detected                 |
+| NEGATIVE 1 | `negativeSanitized`       | `document.location.search`      | `innerHTML` (via `DOMPurify.sanitize`) | NOT detected (barrier)   |
+| NEGATIVE 2 | `negativeHardcoded`       | hardcoded string                | `innerHTML`                            | NOT detected (no source) |
+| EDGE CASE  | `edgeCaseEval`            | `window.location.hash`          | `eval()`                               | Detected                 |
+| NEGATIVE 3 | `negativeEncoded`         | `document.location.search`      | `innerHTML` (via `encodeURIComponent`) | NOT detected (barrier)   |
+
+## Exercises
+
+Each exercise builds on the previous one, progressively teaching more advanced CodeQL concepts.
+
+### Exercise 1: Find XSS Sinks
+
+**Goal**: Write a query that identifies dangerous DOM operations that could introduce XSS.
+
+**Concepts**: `DataFlow::Node`, `DataFlow::PropWrite`, `DataFlow::CallNode`, `DataFlow::globalVarRef`
+
+**What you'll find**:
+
+- Property writes to `innerHTML` and `outerHTML`
+- Calls to `document.write()`
+- Calls to `eval()`
+
+**Validate**:
+
+```bash
+codeql test run exercises-tests/Exercise1
+```
+
+### Exercise 2: Find Remote Flow Sources
+
+**Goal**: Identify all user-controlled inputs that could be XSS sources.
+
+**Concepts**: `RemoteFlowSource`, `getSourceType()`
+
+**What you'll find**:
+
+- `document.location.search`, `window.location.hash`
+- URL parameters via `URLSearchParams`
+- Any other browser API that returns user-controlled data
+
+**Validate**:
+
+```bash
+codeql test run exercises-tests/Exercise2
+```
+
+### Exercise 3: Basic Taint-Tracking Configuration
+
+**Goal**: Connect sources to sinks using CodeQL's taint-tracking framework.
+
+**Concepts**: `DataFlow::ConfigSig`, `TaintTracking::Global`, `isSource`, `isSink`, `flow()`
+
+**Key pattern**:
+
+```ql
+module MyConfig implements DataFlow::ConfigSig {
+  predicate isSource(DataFlow::Node source) { ... }
+  predicate isSink(DataFlow::Node sink) { ... }
+}
+module MyFlow = TaintTracking::Global<MyConfig>;
+```
+
+This exercise will find ALL XSS flows including the sanitized ones (no barriers yet).
+
+**Validate**:
+
+```bash
+codeql test run exercises-tests/Exercise3
+```
+
+### Exercise 4: Add Sanitizer Barriers
+
+**Goal**: Reduce false positives by adding barriers for sanitizer functions.
+
+**Concepts**: `isBarrier` predicate, `DataFlow::CallNode`, `getCalleeName()`
+
+**What changes**: Flows through `DOMPurify.sanitize()` and `encodeURIComponent()` are blocked.
+
+**Validate**:
+
+```bash
+codeql test run exercises-tests/Exercise4
+```
+
+### Exercise 5: Full Path-Problem Query
+
+**Goal**: Convert the query to a production-style `path-problem` query with full taint paths.
+
+**Concepts**: `@kind path-problem`, `PathGraph`, `PathNode`, `flowPath()`
+
+**What changes**:
+
+- Query metadata includes `@kind path-problem`, `@security-severity`, CWE tags
+- Uses `PathNode` instead of `Node` for source/sink
+- Imports `PathGraph` for visualization
+- Shows complete taint propagation paths in results
+
+**Validate**:
+
+```bash
+codeql test run exercises-tests/Exercise5
+```
+
+## Validating Solutions
+
+```bash
+# Run all solution tests (should pass after accepting results)
+codeql test run solutions-tests
+
+# Run a specific exercise test
+codeql test run solutions-tests/Exercise3
+
+# Accept current results as expected baseline
+codeql test run solutions-tests --learn
+```
+
+## Learning Path
+
+```
+Exercise 1 (Sinks)
+    ↓
+Exercise 2 (Sources)
+    ↓
+Exercise 3 (Taint Tracking) ← Core concept
+    ↓
+Exercise 4 (Barriers) ← Reducing false positives
+    ↓
+Exercise 5 (Path Problem) ← Production-ready query
+```
+
+## Relation to Production Query
+
+| Workshop Concept                         | Production Query (`UI5Xss.ql`)               |
+| ---------------------------------------- | -------------------------------------------- |
+| `RemoteFlowSource`                       | `RemoteFlowSource` + SAP-specific sources    |
+| `innerHTML`/`document.write` sinks       | SAP UI5 HTML injection sinks                 |
+| `sanitize`/`encodeURIComponent` barriers | `encodeHTML`/`encodeJS`/`encodeCSS` barriers |
+| `DataFlow::ConfigSig`                    | Same pattern                                 |
+| `path-problem`                           | Same query kind                              |
+
+## Tips
+
+- Start with Exercise 1 and work sequentially — each builds on the previous
+- Use `codeql test run --learn` to accept current output as expected results
+- If stuck, look at the solution for the PREVIOUS exercise as a starting point
+- The `tests-common/test.js` file has comments explaining each test case

.github/skills/create-codeql-query-development-workshop/examples/codeql-sap-js-ui5-xss/build-databases.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+set -e
+
+WORKSHOP_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+
+echo "Building test databases for SAP UI5 XSS workshop..."
+
+# Build solution test databases
+for exercise_dir in "${WORKSHOP_ROOT}"/solutions-tests/Exercise*/; do
+    exercise_name="$(basename "$exercise_dir")"
+    DB_PATH="${exercise_dir}/${exercise_name}.testproj"
+
+    echo "  Creating database: ${exercise_name} (solutions)"
+    rm -rf "${DB_PATH}"
+
+    codeql test extract --search-path="${WORKSHOP_ROOT}/solutions" "$exercise_dir"
+done
+
+# Build exercise test databases
+for exercise_dir in "${WORKSHOP_ROOT}"/exercises-tests/Exercise*/; do
+    exercise_name="$(basename "$exercise_dir")"
+    DB_PATH="${exercise_dir}/${exercise_name}.testproj"
+
+    echo "  Creating database: ${exercise_name} (exercises)"
+    rm -rf "${DB_PATH}"
+
+    codeql test extract --search-path="${WORKSHOP_ROOT}/exercises" "$exercise_dir"
+done
+
+echo "Database creation complete!"
+echo ""
+echo "To run tests:"
+echo "  codeql test run solutions-tests    # Validate solutions"
+echo "  codeql test run exercises-tests    # Test your exercises"

.github/skills/create-codeql-query-development-workshop/examples/codeql-sap-js-ui5-xss/codeql-workspace.yml

@@ -0,0 +1,2 @@
+provide:
+  - '*/codeql-pack.yml'

.github/skills/create-codeql-query-development-workshop/examples/codeql-sap-js-ui5-xss/exercises-tests/Exercise1/Exercise1.expected

@@ -0,0 +1 @@
+Exercise1.ql

.github/skills/create-codeql-query-development-workshop/examples/codeql-sap-js-ui5-xss/exercises-tests/Exercise1/Exercise1.qlref

@@ -0,0 +1,68 @@
+// Test cases for XSS taint-tracking workshop
+// Inspired by SAP UI5 XSS production query patterns
+
+// ============================================================
+// POSITIVE CASE 1: Direct XSS via document.location.search → innerHTML
+// ============================================================
+function positiveDirectInnerHTML() {
+  var userInput = document.location.search; // source: URL query string
+  var container = document.getElementById("output");
+  container.innerHTML = userInput; // sink: innerHTML assignment — XSS!
+}
+
+// ============================================================
+// POSITIVE CASE 2: XSS via window.location.hash → document.write()
+// ============================================================
+function positiveDocumentWrite() {
+  var hashValue = window.location.hash; // source: URL hash fragment
+  document.write(hashValue); // sink: document.write — XSS!
+}
+
+// ============================================================
+// POSITIVE CASE 3: XSS via URL param parsed → outerHTML
+// ============================================================
+function positiveOuterHTML() {
+  var params = new URLSearchParams(window.location.search);
+  var name = params.get("name"); // source: user-controlled URL parameter
+  var el = document.getElementById("profile");
+  el.outerHTML = "<div>" + name + "</div>"; // sink: outerHTML assignment — XSS!
+}
+
+// ============================================================
+// NEGATIVE CASE 1: Sanitized value via DOMPurify.sanitize() → innerHTML
+// Should NOT be flagged because the input is sanitized
+// ============================================================
+function negativeSanitized() {
+  var userInput = document.location.search; // source
+  var clean = DOMPurify.sanitize(userInput); // barrier: sanitization
+  var container = document.getElementById("safe-output");
+  container.innerHTML = clean; // safe: sanitized input
+}
+
+// ============================================================
+// NEGATIVE CASE 2: Hardcoded string → innerHTML
+// Should NOT be flagged because there is no user-controlled source
+// ============================================================
+function negativeHardcoded() {
+  var container = document.getElementById("static-content");
+  container.innerHTML = "<p>Welcome to the application</p>"; // safe: hardcoded
+}
+
+// ============================================================
+// EDGE CASE: XSS via eval() with URL input
+// ============================================================
+function edgeCaseEval() {
+  var code = window.location.hash.substring(1); // source: URL hash
+  eval(code); // sink: eval — code injection / XSS!
+}
+
+// ============================================================
+// NEGATIVE CASE 3: encodeURIComponent barrier
+// Should NOT be flagged because input is encoded
+// ============================================================
+function negativeEncoded() {
+  var userInput = document.location.search;
+  var encoded = encodeURIComponent(userInput); // barrier: encoding
+  var container = document.getElementById("encoded-output");
+  container.innerHTML = encoded; // safe: encoded input
+}