swift-expert
Fail
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: CRITICALINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-defined requirements and platform targets to generate executable Swift code, creating a surface for indirect instructions to influence agent behavior. \n
- Ingestion points: Architecture analysis steps and platform target definitions in SKILL.md. \n
- Boundary markers: None identified; the skill does not use specific delimiters to isolate external requirements from instructions. \n
- Capability inventory: Instructions in SKILL.md trigger local command execution via
swift buildandswift test. \n - Sanitization: No evidence of input validation or sanitization was found. \n- [COMMAND_EXECUTION]: The skill workflow instructs the agent to execute shell commands for building and testing code. \n
- Evidence: Core Workflow validation checkpoints in SKILL.md call for
swift build,swift build -warnings-as-errors, andswift test. \n- [EXTERNAL_DOWNLOADS]: A documentation link provided in the skill metadata is flagged by automated reputation scanners. \n - Evidence: The URL
https://jeffallan.github.io/claude-skills/skills/language/swift-expert/in SKILL.md is identified as malicious by automated scanners.
Recommendations
- Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata