The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of App Store metadata.
Ingestion points: The references/release-runbook.md file specifies that scripts/validate_app_store_metadata.py processes files in fastlane/metadata/en-US.
Boundary markers: The SKILL.md file defines a "Safety Boundary" requiring explicit user confirmation before performing release actions.
Capability inventory: The skill uses xcodebuild (archiving/exporting), fastlane (uploading), and altool (uploading), as well as executing local shell and Python scripts.
Sanitization: There is no explicit mention of sanitization or escaping for the content within the metadata files before they are processed by the validation script.
[COMMAND_EXECUTION]: The skill executes multiple local commands and scripts to perform its primary function.
Evidence: Execution of scripts/ios_preflight.sh, scripts/validate_app_store_metadata.py, xcodebuild, fastlane, and xcrun altool as detailed in references/release-runbook.md.
Context: These are standard operations for an iOS release workflow and are executed locally within the workspace context.

ios-app-store-release