xcode-build
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATION
Full Analysis
- Indirect Prompt Injection (HIGH): The skill processes Xcode workspaces and projects which are untrusted external data sources.
- Ingestion points: Path to
.xcworkspaceor.xcodeprojprovided by user or discovered via Glob. - Boundary markers: Absent; the agent is instructed to use direct CLI tools on these files without validation.
- Capability inventory: The
xcodebuildtool inherently executes arbitrary shell scripts defined within the project's build phases. - Sanitization: Absent; there is no check for malicious scripts in the project file before building.
- Command Execution (HIGH): The skill grants the agent broad access to the
Bashtool to run powerful system utilities likexcodebuildandxcrun, which can be subverted by malicious project configurations to gain persistent access. - Data Exposure (MEDIUM): The use of
/usr/bin/log streamallows the agent to capture real-time system or application logs. - Evidence: The provided patterns for log capture use predicates that could be expanded or bypassed to exfiltrate sensitive information from other system processes.
Recommendations
- AI detected serious security threats
Audit Metadata