The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: This skill is designed to ingest and analyze untrusted iOS source code, pull requests, and project configuration files. It lacks explicit boundary markers or instructions to disregard potential instructions embedded within the reviewed code. The guidelines suggest using build tools such as 'xcodebuild' or repository-specific 'wrapper commands', which creates a capability surface where an attacker could influence the agent's behavior or execute malicious scripts via the build process. Evidence: The 'Tools' and 'UI Test Focus' sections in 'references/current-guidance.md' specify the use of build and test commands on project files. Capability Inventory: The agent is instructed to use subprocesses for 'xcodebuild' and potentially arbitrary 'repo wrapper commands'. Sanitization: There are no instructions for sanitizing or escaping the content of the source code before analysis or execution.

ios-review