The Agent Skills Directory

[SAFE]: The skill provides standard iOS development templates and architectural guidance. No malicious code, obfuscation, or suspicious patterns were detected across the provided files.
[CREDENTIALS_UNSAFE]: No hardcoded credentials or secrets were found. The skill correctly demonstrates using the iOS Keychain for secure token storage in references/networking.md.
[DATA_EXFILTRATION]: Networking logic in references/networking.md uses standard URLSession with developer-defined base URLs. No unauthorized data transmission to external or suspicious domains was observed.
[PROMPT_INJECTION]: No attempts to override agent instructions or bypass safety filters were found. The prompt templates for on-device AI in references/foundation-models.md are well-structured and focused on specific feature tasks.
[INDIRECT_PROMPT_INJECTION]: While the skill defines patterns for processing external data such as deep link URLs and user-provided context for AI prompts, it includes explicit instructions to exclude secrets and PII from prompts, and uses structured boundaries (e.g., 'Context:' and 'Values:' labels) to manage the input surface.

ios-platform