feat: add 3 agent security skills (MCP audit, OWASP compliance, supply chain) by imran-siddique · Pull Request #1248 · github/awesome-copilot

imran-siddique · 2026-04-01T04:10:55Z

Adds 3 new skills for AI agent security and governance — an area not currently covered in the awesome-copilot collection.

New Skills

1. mcp-security-audit

Audit MCP server configurations for security issues:

Hardcoded secrets detection (API keys, tokens, credentials)
Shell injection pattern detection in server args
Unpinned dependency flagging (@latest)
Dangerous command patterns (\�val, \�ash -c, pipe-to-shell)

2. agent-owasp-compliance

Check agent systems against the OWASP Agentic Security Initiative (ASI) Top 10:

Maps all 10 ASI risks with detection signals and passing/failing examples
Generates X/10 compliance report
Includes quick assessment questions for rapid evaluation
Covers risks specific to agents (not LLMs): tool governance, trust boundaries, delegation

3. agent-supply-chain

Supply chain integrity for agent plugins and tools:

SHA-256 integrity manifest generation and verification
Tamper detection (modified, deleted, untracked files)
Dependency version pinning audit
Promotion gate pattern (dev to production readiness check)
CI integration example

Why These Skills

The existing \�gent-governance\ skill covers governance patterns. These 3 skills extend into specific operational areas:

MCP audit: MCP servers are everywhere but ungoverned
OWASP compliance: Industry standard that most agent systems don't measure against
Supply chain: No agent ecosystem has npm-provenance-equivalent integrity verification yet

Related: Agent Governance Toolkit

Copilot

Pull request overview

This PR adds three new security/governance-focused Agent Skills to expand the repository’s coverage of agent security operations: MCP configuration auditing, OWASP ASI-oriented compliance checking, and agent/plugin supply-chain integrity verification.

Changes:

Added a new mcp-security-audit skill describing checks for secrets, shell-injection patterns, and unpinned MCP server dependencies.
Added a new agent-owasp-compliance skill outlining an OWASP ASI Top 10-oriented assessment workflow and report format.
Added a new agent-supply-chain skill with integrity manifest generation/verification patterns and CI gating examples.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 9 comments.

File	Description
skills/mcp-security-audit/SKILL.md	New skill documentation for auditing `.mcp.json` server configs for secrets, injection patterns, and version pinning issues.
skills/agent-owasp-compliance/SKILL.md	New skill documentation for assessing agent systems against OWASP ASI Top 10 risks and producing a compliance report.
skills/agent-supply-chain/SKILL.md	New skill documentation for hashing-based integrity manifests, tamper detection, dependency pinning audits, and CI verification patterns.

Copilot · 2026-04-01T04:16:13Z

skills/mcp-security-audit/SKILL.md

+description: |
+  Audit MCP (Model Context Protocol) server configurations for security issues. Use this skill when:
+  - Reviewing .mcp.json files for security risks
+  - Checking MCP server args for hardcoded secrets or shell injection patterns
+  - Validating that MCP servers use pinned versions (not @latest)
+  - Detecting unpinned dependencies in MCP server configurations
+  - Auditing which MCP servers a project registers and whether they're on an approved list
+  - Checking for environment variable usage vs. hardcoded credentials in MCP configs
+  - Any request like "is my MCP config secure?", "audit my MCP servers", or "check .mcp.json"


These new skills should be added to the skills index in docs/README.skills.md so they show up in the repo’s documented skill catalog/discovery list (add entries for mcp-security-audit, agent-owasp-compliance, and agent-supply-chain).

Suggested change

description: |

Audit MCP (Model Context Protocol) server configurations for security issues. Use this skill when:

- Reviewing .mcp.json files for security risks

- Checking MCP server args for hardcoded secrets or shell injection patterns

- Validating that MCP servers use pinned versions (not @latest)

- Detecting unpinned dependencies in MCP server configurations

- Auditing which MCP servers a project registers and whether they're on an approved list

- Checking for environment variable usage vs. hardcoded credentials in MCP configs

- Any request like "is my MCP config secure?", "audit my MCP servers", or "check .mcp.json"

description: 'audit mcp server configurations in .mcp.json files for security issues including secrets exposure, shell injection, unpinned dependencies, and unapproved servers'

Copilot · 2026-04-01T04:16:13Z

skills/mcp-security-audit/SKILL.md

+    findings = []
+    args_text = json.dumps(server_config.get("args", []))
+    for pattern, description in DANGEROUS_PATTERNS:
+        if re.search(pattern, args_text):
+            findings.append({


This snippet calls json.dumps(...) and re.search(...) but this code block doesn’t include imports for json/re (they’re only present in a previous snippet). Add the imports here or explicitly note that this block depends on earlier imports so it’s copy/pasteable.

Copilot · 2026-04-01T04:16:13Z

skills/mcp-security-audit/SKILL.md

+
+Flag MCP servers using `@latest` or unversioned packages.
+
+```python
+def check_pinned_versions(server_config: dict) -> list[dict]:


The text says this check flags "unversioned packages", but the shown implementation only flags @latest (and otherwise doesn’t emit findings). Either implement an unversioned/unpinned package detection or adjust the wording to match what the code actually checks.

Copilot · 2026-04-01T04:16:13Z

skills/mcp-security-audit/SKILL.md

+            continue
+        findings = []
+        findings.extend(check_secrets(config))
+        findings.extend(check_shell_injection(server_config))
+        findings.extend(check_pinned_versions(server_config))


check_secrets(config) is executed inside the per-server loop, which will duplicate the same secret findings for every server and can misattribute them. Run the secret scan once outside the loop and/or scope it to the specific server config being evaluated.

Copilot · 2026-04-01T04:16:14Z

skills/agent-supply-chain/SKILL.md

+def verify_manifest(plugin_dir: str) -> tuple[bool, list[str]]:
+    """Verify plugin files against INTEGRITY.json."""
+    root = Path(plugin_dir)
+    manifest_path = root / "INTEGRITY.json"
+


The verify_manifest snippet relies on Path, json, hash_file, and generate_manifest, but this code block doesn’t show the needed imports/definitions. Consider adding minimal imports here or adding a brief note that it depends on the previous snippet so readers don’t copy an incomplete block.

Copilot · 2026-04-01T04:16:14Z

skills/agent-supply-chain/SKILL.md

+    required = ["README.md", ".claude-plugin/plugin.json"]
+    missing = [f for f in required if not (root / f).exists()]


promotion_check hard-codes .claude-plugin/plugin.json as a required file. If this is meant to apply to this repo’s plugin layout, the canonical path is .github/plugin/plugin.json (see e.g. plugins/awesome-copilot/.github/plugin/plugin.json). Otherwise, please clarify that this required-files list is specific to Claude-style plugins so it’s not misleading.

Suggested change

required = ["README.md", ".claude-plugin/plugin.json"]

missing = [f for f in required if not (root / f).exists()]

# README is always required; plugin manifest can be in either canonical or legacy location

required = ["README.md"]

missing = [f for f in required if not (root / f).exists()]

# Require at least one plugin manifest in the repo's canonical layout or the legacy Claude layout

plugin_manifest_paths = [

root / ".github/plugin/plugin.json",

root / ".claude-plugin/plugin.json",

]

if not any(p.exists() for p in plugin_manifest_paths):

missing.append(".github/plugin/plugin.json (or .claude-plugin/plugin.json)")

Copilot · 2026-04-01T04:16:14Z

skills/agent-supply-chain/SKILL.md

+            for c in iter(lambda: f.read(8192), b''):
+                h.update(c)
+        return h.hexdigest()
+
+    manifest = json.loads(Path('INTEGRITY.json').read_text())


The CI example reads INTEGRITY.json from the current working directory. Unless the workflow cds into the plugin directory first, this will fail for manifests stored under a plugin subfolder (as shown earlier). Consider updating the snippet to cd into the plugin dir or to reference the manifest path explicitly.

Copilot · 2026-04-01T04:16:14Z

skills/agent-owasp-compliance/SKILL.md

+Codebase → Scan for each ASI control:
+  ASI-01: Prompt Injection Protection
+  ASI-02: Tool Use Governance
+  ASI-03: Agency Boundaries
+  ASI-04: Escalation Controls


The overview implies there are concrete checks for each ASI-01..ASI-10 control, but the document only defines detailed check sections for a subset (01, 02, 05, 07, 09). Either add guidance/check sections for the remaining risks or make it explicit that only partial checks are included so expectations match.

Copilot · 2026-04-01T04:16:14Z

skills/agent-owasp-compliance/SKILL.md

+        "risk": "ASI-01",
+        "name": "Prompt Injection",
+        "status": "pass" if positive_found and not negative_found else "fail",
+        "controls_found": positive_matches,
+        "vulnerabilities": negative_matches,


In this code snippet, positive_found, negative_found, positive_matches, and negative_matches are undefined, so the example as written isn’t runnable. If this is meant as pseudocode, call that out explicitly; otherwise, include a minimal implementation sketch for computing these values (e.g., via grep/ripgrep or AST scanning).

aaronpowell

It looks like you've incorrectly branched from the main branch not staged, and as a result all the materialised plugins are included in this PR.

You can attempt to fix this with a rebase:

git fetch origin staged
git rebase --onto origin/staged origin/main <branch name>
git push --force-with-lease

If that does not resolve it, you can run npm run plugin:clean which will delete the materialised plugins and you can commit that change.

github-actions · 2026-04-04T16:46:21Z

🔍 Skill Validator Results

4 resource(s) checked | ✅ All checks passed

Full output

Found 3 skill(s)
[agent-owasp-compliance] 📊 agent-owasp-compliance: 2,754 BPE tokens [chars/4: 2,970] (standard ~), 22 sections, 7 code blocks
[agent-owasp-compliance]    ⚠  Skill is 2,754 BPE tokens (chars/4 estimate: 2,970) — approaching "comprehensive" range where gains diminish.
[agent-supply-chain] 📊 agent-supply-chain: 2,515 BPE tokens [chars/4: 2,677] (standard ~), 13 sections, 8 code blocks
[agent-supply-chain]    ⚠  Skill is 2,515 BPE tokens (chars/4 estimate: 2,677) — approaching "comprehensive" range where gains diminish.
[agent-supply-chain]    ⚠  No numbered workflow steps — agents follow sequenced procedures more reliably.
[mcp-security-audit] 📊 mcp-security-audit: 2,264 BPE tokens [chars/4: 2,211] (detailed ✓), 9 sections, 11 code blocks
[mcp-security-audit]    ⚠  No numbered workflow steps — agents follow sequenced procedures more reliably.
�[32m✅ All checks passed (3 skill(s))�[0m

…y chain) - mcp-security-audit: Audit .mcp.json files for hardcoded secrets, shell injection, unpinned versions, dangerous command patterns - agent-owasp-compliance: Check agent systems against OWASP ASI 2026 Top 10 risks with compliance report generation - agent-supply-chain: SHA-256 integrity manifests, tamper detection, version pinning audit, promotion gates for agent plugins Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

@latest

1. Added 3 new skills to docs/README.skills.md index 2. Added imports (json, re) to shell injection check snippet 3. Updated unpinned deps wording to match code behavior (@latest only) 4. Moved check_secrets() outside per-server loop to avoid duplicates 5. Added imports note to verify_manifest snippet 6. Updated promotion_check to support both .github/plugin and .claude-plugin layouts 7. Updated CI example to cd into plugin directory before verifying 8. Added check sections for all 10 ASI controls (was missing 03, 04, 06, 08, 10) 9. Made ASI-01 code snippet runnable with actual file scanning implementation Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

imran-siddique requested a review from aaronpowell as a code owner April 1, 2026 04:10

Copilot AI review requested due to automatic review settings April 1, 2026 04:10

Copilot started reviewing on behalf of imran-siddique April 1, 2026 04:11 View session

Copilot AI reviewed Apr 1, 2026

View reviewed changes

aaronpowell requested changes Apr 1, 2026

View reviewed changes

github-actions bot force-pushed the main branch from 6cc49fc to 6090123 Compare April 1, 2026 22:52

github-actions bot requested a review from dvelton as a code owner April 1, 2026 22:52

github-actions bot force-pushed the main branch from 6090123 to 4d2b7f2 Compare April 1, 2026 22:59

aaronpowell changed the base branch from main to staged April 1, 2026 23:00

imran-siddique force-pushed the skills/agt-security-governance branch from 37d2a45 to eb1b932 Compare April 4, 2026 16:45

imran-siddique and others added 2 commits April 4, 2026 11:27

imran-siddique force-pushed the skills/agt-security-governance branch from eb1b932 to d5052e1 Compare April 4, 2026 18:27

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: add 3 agent security skills (MCP audit, OWASP compliance, supply chain)#1248

feat: add 3 agent security skills (MCP audit, OWASP compliance, supply chain)#1248
imran-siddique wants to merge 2 commits intogithub:stagedfrom
imran-siddique:skills/agt-security-governance

imran-siddique commented Apr 1, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 1, 2026

Uh oh!

Copilot AI Apr 1, 2026

Uh oh!

Copilot AI Apr 1, 2026

Uh oh!

Copilot AI Apr 1, 2026

Uh oh!

Copilot AI Apr 1, 2026

Uh oh!

Copilot AI Apr 1, 2026

Uh oh!

Copilot AI Apr 1, 2026

Uh oh!

Copilot AI Apr 1, 2026

Uh oh!

Copilot AI Apr 1, 2026

Uh oh!

aaronpowell left a comment

Uh oh!

github-actions bot commented Apr 4, 2026 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

		required = ["README.md", ".claude-plugin/plugin.json"]
		missing = [f for f in required if not (root / f).exists()]

-    required = ["README.md", ".claude-plugin/plugin.json"]
-    missing = [f for f in required if not (root / f).exists()]
+    # README is always required; plugin manifest can be in either canonical or legacy location
+    required = ["README.md"]
+    missing = [f for f in required if not (root / f).exists()]
+    # Require at least one plugin manifest in the repo's canonical layout or the legacy Claude layout
+    plugin_manifest_paths = [
+        root / ".github/plugin/plugin.json",
+        root / ".claude-plugin/plugin.json",
+    ]
+    if not any(p.exists() for p in plugin_manifest_paths):
+        missing.append(".github/plugin/plugin.json (or .claude-plugin/plugin.json)")

Conversation

imran-siddique commented Apr 1, 2026

New Skills

1. mcp-security-audit

2. agent-owasp-compliance

3. agent-supply-chain

Why These Skills

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 1, 2026

Choose a reason for hiding this comment

Uh oh!

aaronpowell left a comment

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Apr 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔍 Skill Validator Results

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

github-actions bot commented Apr 4, 2026 •

edited

Loading