The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill instructs the agent to execute chmod +x on the generated scripts/run-uitests.sh file to make it executable. While standard for scaffolding tools, this is an automated privilege escalation on dynamically created content.
[INDIRECT_PROMPT_INJECTION]: The skill audits the project by reading local .swift source files using grep patterns to build a view graph and suggest accessibility identifiers. This creates an attack surface where maliciously crafted source code could influence the agent's logic or output, though the risk is mitigated by specific regex boundary patterns.
[COMMAND_EXECUTION]: The skill generates a shell script that performs complex operations including xcodebuild test and xcrun simctl / xcresulttool. It also programmatically modifies xcodegen.yml or project.yml configuration files to register new test targets.
[DATA_EXPOSURE]: The generated runner script interacts with local developer tools and simulators, potentially exposing details about the local development environment (simulator IDs, build paths) if the script's output is shared without review.

ios-e2e-scaffold