Versioned security assessment

Report ID: SA-D2A32EA0

7/5/2026, 11:03:26 PM

microsoft-foundry security assessment v3

Skill Security Certification Report

Audit History

Audit model: codex Historical report

Skill name: microsoft-foundry
Version: v3
Maintainer: microsoft
Source: github.com/microsoft/github-copilot-for-azure/tree/main/plugin/skills/microsoft-foundry/
Coverage: 22 Files scanned · 3,158 Lines analyzed
Policy version: Unavailable

Highest confirmed finding severity

Medium

3 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.

Static matches were mostly documentation false positives from Markdown, Azure CLI examples, official Microsoft endpoints, and normal configuration guidance. No prompt injection attempt, covert data exfiltration, or malicious override instruction was found. Semantic review identified medium-risk operational concerns around credential output, non-interactive environment propagation, and an unpinned npx package sample.

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

22 Files scanned · 3,158 Lines analyzed

3 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 5 evidence locations

Network access

May connect to external services.

Observed in 25 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 16 evidence locations

External commands

May invoke commands or programs outside the Skill.

Observed in 61 evidence locations

Risk findings

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

Confirmed security concerns (3)

RISK-001 Medium

resource/create/create-foundry-resource.md:109-111 resource/create/references/workflows.md:163-168 rbac/rbac.md:112-116

Credential Output Handling Requires Safeguards

The resource and RBAC workflows show commands that can print account keys and service principal passwords. These are legitimate admin tasks, but outputs must be masked and stored outside chat transcripts.

The cited commands directly retrieve account keys or create service principals with password output. The files include secure-storage guidance, so the risk is operational rather than malicious.

RISK-002 Medium

foundry-agent/deploy/deploy.md:68-72 foundry-agent/deploy/deploy.md:276-278

Non-Interactive Deployment Can Bypass Environment Review

The deployment workflow normally confirms environment variables, but YOLO or non-interactive mode uses resolved values without confirmation. This can deploy unintended sensitive configuration into hosted agent payloads.

The file explicitly says environment variables are included in the agent payload and later says non-interactive mode skips confirmation. The risk depends on the values present in azd or project defaults.

RISK-003 Medium

foundry-agent/create/agent-framework/references/agent-samples.md:27-31

Unpinned Runtime Package Execution In Sample

The Agent Framework sample configures an MCP stdio tool to run npx with @playwright/mcp@latest. Using a floating latest package can introduce supply-chain risk when copied into generated projects.

The sample directly names npx and a latest-tag package. The code is illustrative, but generated or copied use would fetch code whose version is not fixed.

Remediation

Suggested fixes recorded by this audit. Applying them is the maintainer’s responsibility.

FIX-001
Medium
Commands can display account keys or service principal passwords.

Mask credential output by default, avoid placing secrets in chat, and direct users to Key Vault or a CI/CD secret store.
FIX-002
Medium
Non-interactive deployment skips environment variable confirmation.

Require an allowlist for environment variables and keep a confirmation step for any value that appears secret or credential-like.
FIX-003
Medium
The Playwright MCP sample uses an @latest package reference.

Pin the MCP package version and ask for confirmation before adding runtime package execution to generated projects.
FIX-004
Medium
RBAC and provisioning workflows can change cloud access and billing state.

Use least-privilege roles, show the target scope before changes, and require explicit user approval for role assignment or deletion commands.

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 (1)

High

foundry-agent/create/agent-framework/references/agent-as-server.md:67

LLM role tokens injection

if message.role == Role.ASSISTANT:

Force-confirmed blocker/high static finding; AI dismissal overridden.

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