xcode-project-setup
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill requires the agent to execute a bundled Swift command-line tool using
swift run. This tool programmatically updates the local Xcode project files (.xcodeproj) to add dependencies and configuration. - [EXTERNAL_DOWNLOADS]: During execution, the Swift Package Manager fetches the
XcodeProjlibrary and its dependencies (AEXML, PathKit, Spectre) from their respective repositories on GitHub. These are well-established open-source libraries used for project manipulation and are sourced from reputable community organizations. - [INDIRECT_PROMPT_INJECTION]: The tool accepts external parameters such as repository URLs and product names, which are then written into the
.pbxprojfile structure. - Ingestion points: Inputs are parsed from command-line arguments in
scripts/xcode_spm_setup/Sources/main.swift. - Boundary markers: The instructions define a strict command signature in
SKILL.mdbut do not include explicit validation steps to verify the safety of the provided URLs. - Capability inventory: The tool has the capability to write to the local filesystem (modifying project files) and add shell script build phases to the Xcode project.
- Sanitization: The tool utilizes the
XcodeProjlibrary to ensure the project file maintains a valid structure, preventing simple injection attacks that would corrupt the file format. - [PROMPT_INJECTION]: The instructions in
SKILL.mduse strong, imperative language and include a fictional "hefty fee" threat to enforce compliance with operational rules, such as the prohibition against using Ruby. This style of instruction is intended to ensure the agent uses the provided Swift tool rather than attempting alternative, less reliable methods.
Audit Metadata