The Agent Skills Directory

[PROMPT_INJECTION]: The skill generates infrastructure for fetching remote configurations, creating a surface for indirect prompt injection. A compromised remote server could provide malicious data that influences application behavior via the generated flag toggles.
Ingestion points: The generated RemoteFeatureFlagProvider.swift template fetches JSON configuration from a remote URL.
Boundary markers: Data ingestion is constrained by a predefined FeatureFlag enum and extracted via typed getter methods (e.g., boolValue, stringValue).
Capability inventory: The generated infrastructure includes network access via URLSession and local file writes for disk caching.
Sanitization: The code utilizes standard JSONSerialization and JSONDecoder for data validation and structure enforcement.
[DATA_EXFILTRATION]: The generated code includes networking capabilities to fetch data from remote endpoints and file system access to cache configuration values locally.
The RemoteFeatureFlagProvider.swift template performs network requests using URLSession to communicate with user-defined servers.
Configuration data is persisted to a feature_flags_cache.json file within the application's cache directory.

feature-flags