Versioned security assessment

Report ID: SA-554CFFFA

6/30/2026, 7:32:03 AM

azure-prepare security assessment v2

Skill Security Certification Report

Audit model: codex Historical report

Skill name: azure-prepare
Version: v2
Maintainer: microsoft
Source: github.com/microsoft/github-copilot-for-azure/tree/main/plugin/skills/azure-prepare/
Coverage: 183 Files scanned · 18,434 Lines analyzed
Policy version: Unavailable

Highest confirmed finding severity

Critical

2 confirmed security findings require attention.

Installation context

Historical evidence

This report may not describe the currently installable artifact. Open the current Skill page for install guidance.

Open current Skill page

This report does not block or authorize the manifest or ZIP.

The audit confirms a prompt injection attempt in SKILL.md that tells the model to ignore prior training and treat the skill as authoritative. Most static command, filesystem, script, and credential findings are markdown examples or Azure configuration guidance, but the injection wording makes the skill unsafe to publish without remediation.

Report position

Historical report

Open audit history before using this report to install.

Audit attestation

Not attestable

The required immutable binding is incomplete.

Human verification

Not verified

No human verification is recorded for this report.

Coverage

183 Files scanned · 18,434 Lines analyzed

5 items shown for review

Limitations

This report does not claim runtime or sandbox execution and does not prove the absence of side effects.

Evidence chain

Follow the evidence from source binding to the install contract. Available evidence supports verification; it is not a safety guarantee.

Source

Binding unavailable
Artifact

Identity incomplete
Audit

Complete
Install contract

Open manifest to verify
Open manifest

Capabilities observed

Observed means this report recorded supporting evidence. Not recorded does not prove that a capability is absent.

Contains scripts

May execute code included with the Skill.

Observed in 3 evidence locations

Network access

May connect to external services.

Observed in 3 evidence locations

Filesystem access

May read or write local files.

Observed in 3 evidence locations

Env variables

May read values from the process environment.

Observed in 3 evidence locations

External commands

May invoke commands or programs outside the Skill.

Observed in 3 evidence locations

Capability review items (3)

High

references/sdk/azd-deployment.md:7-8

Remote Install Script Piped to Shell

The Azure Developer CLI reference recommends piping a remote installer directly to bash. This is a high-risk execution pattern if copied or run by an agent.

The command is explicitly present and would execute network-fetched code. The context is a legitimate installer, so intent is not clearly malicious.

Medium

SKILL.md:29-35 SKILL.md:81-95

Agent-Driven File Generation Workflow

The skill instructs agents to create plans, infrastructure, Dockerfiles, and Azure configuration files. This is expected but can alter project state.

The file-writing workflow is explicit and central to the skill. User approval is required, which reduces but does not remove the risk.

Medium

references/services/container-apps/environment.md:18-61

Sensitive Configuration Handling Guidance

The references include Azure client IDs, database URLs, and Key Vault secret references. The guidance is mostly safe but involves sensitive deployment configuration.

The examples use Key Vault and managed identity, which are safe patterns. The risk remains because agents may generate or modify secret-related settings.

Risk findings

Confirmed security concerns are separated from items that still need review.

Confirmed security concerns (2)

RISK-001 Critical

Prompt Injection Attempt Detected

SKILL.md claims authoritative control and tells the model to ignore prior training. This can override marketplace safety and user instructions.

The exact file text contains mandatory authority claims and the word IGNORE directed at prior instructions. This matches the prompt injection criteria with direct evidence.

RISK-002 Low

references/runtimes/nodejs.md:183-190

Browser Credential Finding Is Cookie Documentation

The browser credential finding points to a Node.js troubleshooting heading about cookies not setting. This is not credential file access.

The reviewed text is a troubleshooting section for application session cookies. It does not access browser credential files.

Expert evidence

Immutable subject identity, scanner metadata, dismissed matches, and source-level evidence.

Artifact subject

Marketplace commit: Unavailable
Content hash: Unavailable
Tree hash: Unavailable
Skill path: Unavailable
Audit payload hash: Unavailable

Analysis metadata

Audit model: codex

Analysis state: Complete

Scope is limited to the recorded files, lines, methods, and evidence. No runtime or sandbox execution is claimed.

Static false positives ignored (2)

Low

references/analyze.md:17-27 references/recipes/azd/README.md:28-36

Markdown Code Fence Command False Positives

Many external command detections come from markdown code fences, decision trees, or example commands rather than executable skill code.

Targeted review shows these locations are documentation examples. They are not runnable code inside the skill itself.

Low

references/recipes/azd/README.md:42-46 references/plan-template.md:123-130

Relative Path Documentation False Positives

Path traversal detections are mostly relative documentation links and output path examples, not evidence of unsafe path handling.

Reviewed examples are documentation paths such as linked references and intended generated files. No unsafe path join logic was found in these locations.

Verify and export

JSON report CITATION.cff SARIF 2.1.0 Signed installation manifest

The manifest and lockfile bind install artifacts to cryptographic hashes. This integrity claim is separate from the security assessment.

Audit attestation: not_attestable