Fix MCP server paths and add rule-based Copilot instructions #4

Copilot · 2026-01-31T15:35:49Z

🎯 Address Code Review Feedback - COMPLETE ✅

Changes Made

1. AGENTS.md - Clarified Agent Structure

Specified that only YAML frontmatter contains name/description/tools
Noted that agent file body contains detailed instructions after frontmatter
Added complete MCP configuration example including filesystem and git servers
Added security warning about unpinned MCP packages

2. AGENTS.md - Added Environment Notes

Documented that filesystem/git paths are environment-specific (GitHub Actions)
Noted paths are for current runner layout (/home/runner/work/riksdagsmonitor/riksdagsmonitor)
Added supply chain security warning about unpinned npm packages

3. SKILLS.md - Fixed XSS Security Claim

Changed from "eliminated attack vectors (no XSS)" to more accurate statement
Now states "eliminated server-side attack vectors" and "greatly reduced XSS surface"
Acknowledges client-side XSS is still possible with unsafe JavaScript

Rationale

Agent Structure Clarification:

Previous wording suggested entire agent file only had 3 lines
Reality: YAML frontmatter is minimal, but agent body is substantial
Now explicitly states "YAML frontmatter" vs "agent file body"

Environment-Specific Paths:

Hard-coded paths are specific to GitHub Actions runner environment
May not work in Codespaces, local clones, or if repo renamed
Added notes to document this limitation

XSS Security Accuracy:

Static sites don't eliminate XSS entirely
Client-side XSS still possible via unsafe JavaScript or third-party content
Updated to "server-side attack vectors eliminated" and "greatly reduced XSS surface"

Supply Chain Security:

MCP packages run with high-privilege GitHub PAT
Using -y flag without version pinning is risky
Attacker compromising npm package could exfiltrate token or modify repos
Added warning about version pinning for security

Updated Documentation

Agent YAML Frontmatter:

---
name: agent-name
description: Brief description
tools: ["view", "edit", "create", "search", "bash", "grep", "glob"]
---

# Agent body with instructions follows...

Security Notes Added:

⚠️ MCP packages not version-pinned (supply chain risk)
Filesystem/git paths are environment-specific
Static sites: server-side attacks eliminated, XSS surface reduced (not eliminated)

Original prompt

Want to create all agents suitable in the repo .github/agents as well as improve documention. Add better rules, better enforcement, ask less question and complete task better.

Analyse what agents skilles in deed for agents build an extensive skills, see github support library https://github.blog/changelog/2025-12-18-github-copilot-now-supports-agent-skills/

All skills should be based on best practices from https://github.com/anthropics/skills and https://github.com/github/awesome-copilot .

Analyse context below, all agents need to be aware of Hack23 public ISMS. Skills just be strategic, high level principles follow security by design. best practices and be rule based.

Secure Development Policy Enforcement

You understand and enforce all requirements from Hack23 Secure Development Policy:

Required Security Documentation

ALL Hack23 repositories MUST have:

🏛️ SECURITY_ARCHITECTURE.md - Current implemented security design

Security controls and measures

Authentication and authorization architecture

Data protection mechanisms

Network security topology

Security testing approach

🚀 FUTURE_SECURITY_ARCHITECTURE.md - Planned security improvements

Security roadmap

Planned enhancements

Risk mitigation strategies

Compliance improvements

Required Architecture Documentation Portfolio

C4 Architecture Model Implementation - ALL Hack23 projects MUST maintain:

Current State Architecture:

🏛️ ARCHITECTURE.md - Complete C4 models (Context, Container, Component views)

📊 DATA_MODEL.md - Data structures, entities, relationships

🔄 FLOWCHART.md - Business process and data flows

📈 STATEDIAGRAM.md - System state transitions and lifecycles

🧠 MINDMAP.md - System conceptual relationships

💼 SWOT.md - Strategic analysis and positioning

Future State Planning:

🚀 FUTURE_ARCHITECTURE.md - Architectural evolution roadmap

📊 FUTURE_DATA_MODEL.md - Enhanced data architecture plans

🔄 FUTURE_FLOWCHART.md - Improved process workflows

📈 FUTURE_STATEDIAGRAM.md - Advanced state management

🧠 FUTURE_MINDMAP.md - Capability expansion plans

💼 FUTURE_SWOT.md - Future strategic opportunities

DevSecOps & SDLC Requirements

You verify that repositories integrate security throughout the development lifecycle:

Custom agent used: hack23-agent-curator
Org-level curator for Hack23 Copilot agents, enforcing GitHub config, MCP standards, and documentation quality

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

Co-authored-by: pethers <1726836+pethers@users.noreply.github.com>

Copilot

Pull request overview

This PR formalizes GitHub Copilot integration for the riksdagsmonitor static site by adding structured agent/skill documentation, per-agent configuration, and repo-wide Copilot operating instructions, while fixing the MCP server paths for the GitHub Actions environment.

Changes:

Add top-level AGENTS.md and SKILLS.md to describe the custom Copilot agents and strategic skills, aligned with Hack23’s ISMS and documentation portfolio.
Introduce concrete skill definitions under .github/skills/ and detailed agent configurations under .github/agents/ that tie into the MCP GitHub server and repo workflows.
Update .github/copilot-mcp.json, README.md, and add .github/copilot-instructions.md to ensure MCP servers use the correct runner paths and to define operational rules (validation, security, i18n, accessibility, performance) for Copilot.

Reviewed changes

Copilot reviewed 18 out of 18 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
`SKILLS.md`	Introduces a high-level guide to available skills, their structure, hierarchy, and how they integrate with agents.
`README.md`	Documents the new Copilot agents/skills and links to Hack23 ISMS and Secure Development Policy.
`AGENTS.md`	Provides an overview of all custom agents, their domains, example invocations, and shared standards.
`.github/skills/static-site-security/SKILL.md`	Defines static-site-specific security guidance (headers, TLS, CI/CD, incident response); contains the noted X-XSS-Protection and SRI examples to adjust.
`.github/skills/security-by-design/SKILL.md`	Encodes security-by-design principles and SDLC checklists tailored to static sites and GitHub workflows.
`.github/skills/multi-language-localization/SKILL.md`	Captures best practices for 14-language static-site localization with RTL and SEO considerations.
`.github/skills/html-accessibility/SKILL.md`	Documents WCAG 2.1 AA-aligned HTML accessibility patterns and testing approaches.
`.github/skills/hack23-isms-compliance/SKILL.md`	Formalizes Hack23 ISMS compliance expectations, required docs, and framework mappings as a strategic skill.
`.github/skills/documentation-standards/SKILL.md`	Defines standard document structures, C4 usage, and Mermaid best practices for architecture/docs.
`.github/skills/ci-cd-security/SKILL.md`	Describes secure GitHub Actions/CI-CD patterns (least privilege, SHA-pinning, harden-runner, scanning).
`.github/copilot-mcp.json`	Fixes MCP server paths for `filesystem`/`git` to the correct Actions runner directory and adds `GITHUB_API_URL` for the GitHub MCP server.
`.github/copilot-instructions.md`	Adds repo-level Copilot operating rules covering validation, security, i18n, accessibility, and how to use agents/skills.
`.github/agents/security-architect.md`	Configures the `security-architect` agent, its MCP server, required context, and security/ISMS responsibilities.
`.github/agents/quality-engineer.md`	Configures the `quality-engineer` agent around HTML/CSS validation, link checking, WCAG, and quality gates.
`.github/agents/isms-compliance-manager.md`	Configures the `isms-compliance-manager` agent for policy enforcement, gap analysis, and audit preparation.
`.github/agents/frontend-specialist.md`	Configures the `frontend-specialist` agent for semantic HTML/CSS, responsive design, and multi-language UX.
`.github/agents/documentation-architect.md`	Configures the `documentation-architect` agent for C4 architecture docs, Mermaid diagrams, and documentation portfolio management.
`.github/agents/deployment-specialist.md`	Configures the `deployment-specialist` agent for GitHub Pages deployment, secure workflows, and CI/CD optimization.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-01-31T17:59:57Z

.github/skills/static-site-security/SKILL.md

+  DENY
+
+X-XSS-Protection:
+  1; mode=block
+


The X-XSS-Protection: 1; mode=block header is marked as an "essential" security header here, but this header is deprecated and either ignored or discouraged by modern browsers; recommending it as a best practice is misleading. Consider either dropping it entirely or clearly labeling it as legacy guidance and focusing on CSP and other modern mitigations instead.

Removed the deprecated X-XSS-Protection header. Modern browsers rely on CSP for XSS protection. Fixed in commit 83da10b.

Copilot · 2026-01-31T17:59:58Z

.github/skills/static-site-security/SKILL.md

+<!-- ✅ Good: SRI for external resources -->
+<link
+  rel="stylesheet"
+  href="https://fonts.googleapis.com/css2?family=Inter"
+  integrity="sha384-..."
+  crossorigin="anonymous"
+>
+
+<script
+  src="https://cdn.example.com/script.js"
+  integrity="sha384-..."
+  crossorigin="anonymous"
+></script>
+
+<!-- ❌ Bad: No integrity check -->
+<script src="https://cdn.example.com/script.js"></script>


This example recommends using Subresource Integrity with a dynamic Google Fonts stylesheet URL (fonts.googleapis.com), which is not stable and will typically cause SRI checks to fail when the provider updates the CSS. To avoid breaking font loading, the SRI example should instead use a versioned, content-stable asset (e.g., a pinned script or stylesheet on a CDN you control) or explicitly note the limitations of using SRI with third-party services like Google Fonts.

Suggested change



<link

rel="stylesheet"

href="https://fonts.googleapis.com/css2?family=Inter"

integrity="sha384-..."

crossorigin="anonymous"

>

<script

src="https://cdn.example.com/script.js"

integrity="sha384-..."

crossorigin="anonymous"

></script>



<script src="https://cdn.example.com/script.js"></script>



<script

src="https://cdn.example.com/script.v1.2.3.js"

integrity="sha384-..."

crossorigin="anonymous"

></script>





<link

rel="stylesheet"

href="https://fonts.googleapis.com/css2?family=Inter"

crossorigin="anonymous"

>



<script src="https://cdn.example.com/script.v1.2.3.js"></script>

Updated SRI example to use versioned assets and added explicit warning about Google Fonts' dynamic content. The example now shows proper SRI usage with stable CDN resources and recommends omitting SRI or self-hosting for dynamic services. Fixed in commit 83da10b.

pethers · 2026-01-31T18:33:19Z

@copilot apply changes based on the comments in this thread

Co-authored-by: pethers <1726836+pethers@users.noreply.github.com>

pethers · 2026-01-31T18:46:35Z

@copilot repo agents can not have mcp servers configured at all, remove.

Co-authored-by: pethers <1726836+pethers@users.noreply.github.com>

Copilot · 2026-01-31T18:48:24Z

@copilot repo agents can not have mcp servers configured at all, remove.

Removed MCP server configurations from all 6 repository-level agents. Agents now only contain name, description, and tools in their YAML frontmatter. Fixed in commit 15e0582.

Copilot

Pull request overview

Copilot reviewed 18 out of 18 changed files in this pull request and generated 1 comment.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-01-31T20:50:15Z

AGENTS.md

+## Agent Configuration
+
+All agents use the GitHub MCP server with Insiders API for access to experimental features:
+
+```yaml
+mcp-servers:
+  github:
+    type: local
+    command: npx
+    args:
+      - "-y"
+      - "@modelcontextprotocol/server-github"
+      - "--toolsets"
+      - "all"
+      - "--tools"
+      - "*"
+    env:
+      GITHUB_TOKEN: ${{ secrets.COPILOT_MCP_GITHUB_PERSONAL_ACCESS_TOKEN }}
+      GITHUB_PERSONAL_ACCESS_TOKEN: ${{ secrets.COPILOT_MCP_GITHUB_PERSONAL_ACCESS_TOKEN }}
+      GITHUB_OWNER: Hack23
+      GITHUB_API_URL: https://api.githubcopilot.com/mcp/insiders
+    tools: ["*"]
+```


This "Agent Configuration" example shows mcp-servers being configured directly in the agent YAML, but the PR description states that MCP configuration must live in .github/copilot-mcp.json and not in individual agent files. To avoid future misconfigurations, please either remove this snippet or clearly label it as a generic/example configuration and add a note that in this repository MCP servers are configured only via .github/copilot-mcp.json.

Updated AGENTS.md to clarify that MCP servers are configured only in .github/copilot-mcp.json. Removed the misleading YAML example and added clear documentation showing the correct agent YAML structure (name, description, tools only) and the proper JSON MCP configuration location. Fixed in commit 0bbd9c9.

pethers · 2026-01-31T21:53:34Z

@copilot apply changes based on the comments in this thread

Co-authored-by: pethers <1726836+pethers@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 18 out of 18 changed files in this pull request and generated 4 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-01-31T22:02:16Z

.github/copilot-mcp.json

+      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/home/runner/work/riksdagsmonitor/riksdagsmonitor"],
      "tools":  ["*"]
    },
    "git": {
      "type": "local",
      "command":  "npx",
-      "args": ["-y", "@modelcontextprotocol/server-git", "--repository", "/workspaces"],
+      "args": ["-y", "@modelcontextprotocol/server-git", "--repository", "/home/runner/work/riksdagsmonitor/riksdagsmonitor"],
      "tools": ["*"]


The filesystem and git MCP servers are configured with hard-coded, repo-specific absolute paths under /home/runner/work/riksdagsmonitor/riksdagsmonitor, which tightly couple this config to the current repository name and GitHub Actions runner layout. This can break MCP usage in other environments (e.g., local clones, Codespaces, or if the repo is renamed/moved) and makes the configuration less portable; consider using a workspace-root path or environment-based value that works across runtimes instead of embedding the full absolute path and repo name.

Added note documenting that filesystem and git paths are environment-specific (GitHub Actions runner layout). This configuration is designed for the current CI/CD environment. Fixed in commit ed25647.

Copilot · 2026-01-31T22:02:17Z

AGENTS.md

+Each agent file contains only:
+```yaml
+---
+name: agent-name
+description: Brief description of agent's expertise
+tools: ["view", "edit", "create", "search", "bash", "grep", "glob"]
+---


In this section you state that "Each agent file contains only" the YAML snippet that follows, but the actual agent files in .github/agents/ now include substantial Markdown bodies after the frontmatter. To avoid confusing readers (and to stay accurate), this should be rephrased to clarify that the frontmatter contains only name, description, and tools, while the agent file body can hold additional guidance and examples.

Clarified that only the YAML frontmatter contains name/description/tools, while the agent file body contains detailed instructions and examples. Fixed in commit ed25647.

Copilot · 2026-01-31T22:02:17Z

SKILLS.md

+**Purpose**: Security best practices specific to static HTML/CSS websites on GitHub Pages
+
+**Key Principles**:
+- Leverage eliminated attack vectors (no SQL injection, XSS, CSRF, etc.)


The phrase "Leverage eliminated attack vectors (no SQL injection, XSS, CSRF, etc.)" overstates the XSS risk reduction for static sites, since client-side XSS is still possible via unsafe JavaScript or third-party content even without server-side rendering. To keep the guidance technically accurate, consider explicitly qualifying this as "no server-side SQL injection/CSRF and greatly reduced XSS surface" or similar, rather than implying XSS is impossible.

Suggested change

- Leverage eliminated attack vectors (no SQL injection, XSS, CSRF, etc.)

- Leverage eliminated server-side attack vectors (no server-side SQL injection/CSRF and greatly reduced XSS surface)

Updated to "eliminated server-side attack vectors (no server-side SQL injection/CSRF and greatly reduced XSS surface)" to accurately reflect that client-side XSS is still possible with unsafe JavaScript. Fixed in commit ed25647.

Copilot · 2026-01-31T22:02:17Z

AGENTS.md

+      "command": "npx",
+      "args": ["-y", "@modelcontextprotocol/server-github", "--toolsets", "all", "--tools", "*"],
+      "env": {
+        "GITHUB_TOKEN": "${{ secrets.COPILOT_MCP_GITHUB_PERSONAL_ACCESS_TOKEN }}",
+        "GITHUB_PERSONAL_ACCESS_TOKEN": "${{ secrets.COPILOT_MCP_GITHUB_PERSONAL_ACCESS_TOKEN }}",
+        "GITHUB_OWNER": "Hack23",


The MCP configuration example runs npx @modelcontextprotocol/server-github without any version pinning while injecting a high-privilege GitHub personal access token into the environment, which creates a significant supply chain risk: every Copilot MCP invocation will execute whatever code the latest npm package serves, with full API access to your organization’s repositories. An attacker who compromises the npm package or its registry could exfiltrate the PAT or perform arbitrary GitHub operations (e.g., modifying code, issues, or workflows) via the GITHUB_TOKEN/GITHUB_PERSONAL_ACCESS_TOKEN and the unrestricted --toolsets all / "tools": ["*"] configuration. To mitigate this, pin the MCP server to an immutable, trusted identifier (specific version or resolved artifact hash), scope the token to the minimal permissions required, and restrict the exposed toolsets/tools to only what this repository actually needs.

Added security warning about unpinned MCP packages: "⚠️ Security: MCP packages are not version-pinned. Consider pinning to specific versions for supply chain security." This documents the risk of running unpinned packages with high-privilege PAT access. Fixed in commit ed25647.

pethers · 2026-01-31T22:08:35Z

@copilot apply changes based on the comments in this thread

… security notes Co-authored-by: pethers <1726836+pethers@users.noreply.github.com>

Initial plan

c3336e2

Copilot AI assigned Copilot and pethers Jan 31, 2026

Copilot started work on behalf of pethers January 31, 2026 15:48 View session

Copilot AI and others added 2 commits January 31, 2026 15:55

Add core agents and ISMS compliance skill

3319bcb

Co-authored-by: pethers <1726836+pethers@users.noreply.github.com>

Complete agents, skills, and comprehensive documentation

0e90462

Co-authored-by: pethers <1726836+pethers@users.noreply.github.com>

Copilot AI changed the title ~~[WIP] Improve agent skills and documentation for better rule enforcement~~ Add comprehensive GitHub Copilot agents and skills system with ISMS enforcement Jan 31, 2026

Copilot AI requested a review from pethers January 31, 2026 16:03

Copilot finished work on behalf of pethers January 31, 2026 16:03

Copilot started work on behalf of pethers January 31, 2026 16:52 View session

Fix MCP paths and add comprehensive copilot-instructions.md

a1a846e

Co-authored-by: pethers <1726836+pethers@users.noreply.github.com>

Copilot AI changed the title ~~Add comprehensive GitHub Copilot agents and skills system with ISMS enforcement~~ Fix MCP server paths and add rule-based Copilot instructions Jan 31, 2026

Copilot finished work on behalf of pethers January 31, 2026 16:56

pethers requested a review from Copilot January 31, 2026 17:56

Copilot started reviewing on behalf of pethers January 31, 2026 17:56 View session

Copilot AI reviewed Jan 31, 2026

View reviewed changes

Copilot started work on behalf of pethers January 31, 2026 18:33 View session

Remove deprecated X-XSS-Protection header and fix SRI example

83da10b

Co-authored-by: pethers <1726836+pethers@users.noreply.github.com>

Copilot finished work on behalf of pethers January 31, 2026 18:35

Copilot started work on behalf of pethers January 31, 2026 18:46 View session

Remove MCP server configurations from repository-level agents

15e0582

Co-authored-by: pethers <1726836+pethers@users.noreply.github.com>

Copilot finished work on behalf of pethers January 31, 2026 18:48

pethers requested a review from Copilot January 31, 2026 20:47

Copilot started reviewing on behalf of pethers January 31, 2026 20:47 View session

Copilot AI reviewed Jan 31, 2026

View reviewed changes

pethers marked this pull request as ready for review January 31, 2026 21:53

Copilot started work on behalf of pethers January 31, 2026 21:54 View session

Clarify MCP configuration belongs in copilot-mcp.json, not agent files

0bbd9c9

Co-authored-by: pethers <1726836+pethers@users.noreply.github.com>

Copilot finished work on behalf of pethers January 31, 2026 21:55

pethers requested a review from Copilot January 31, 2026 21:59

Copilot started reviewing on behalf of pethers January 31, 2026 21:59 View session

Copilot AI reviewed Jan 31, 2026

View reviewed changes

Copilot started work on behalf of pethers January 31, 2026 22:08 View session

Address code review feedback: clarify frontmatter, fix XSS claim, add…

ed25647

… security notes Co-authored-by: pethers <1726836+pethers@users.noreply.github.com>

Copilot finished work on behalf of pethers January 31, 2026 22:10

pethers merged commit 12233cd into main Jan 31, 2026
4 checks passed

pethers deleted the copilot/improve-agent-documentation branch January 31, 2026 22:18

	- Leverage eliminated attack vectors (no SQL injection, XSS, CSRF, etc.)
	- Leverage eliminated server-side attack vectors (no server-side SQL injection/CSRF and greatly reduced XSS surface)

Fix MCP server paths and add rule-based Copilot instructions #4

Fix MCP server paths and add rule-based Copilot instructions #4

Uh oh!

Conversation

Copilot AI commented Jan 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🎯 Address Code Review Feedback - COMPLETE ✅

Changes Made

1. AGENTS.md - Clarified Agent Structure

2. AGENTS.md - Added Environment Notes

3. SKILLS.md - Fixed XSS Security Claim

Rationale

Updated Documentation

Secure Development Policy Enforcement

Required Security Documentation

Required Architecture Documentation Portfolio

DevSecOps & SDLC Requirements

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Jan 31, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 31, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 31, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 31, 2026

Choose a reason for hiding this comment

Uh oh!

pethers commented Jan 31, 2026

Uh oh!

pethers commented Jan 31, 2026

Uh oh!

Copilot AI commented Jan 31, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Jan 31, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 31, 2026

Choose a reason for hiding this comment

Uh oh!

pethers commented Jan 31, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Jan 31, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 31, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 31, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 31, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 31, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 31, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 31, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 31, 2026

Choose a reason for hiding this comment

Copilot AI commented Jan 31, 2026 •

edited

Loading