The Agent Skills Directory

[SAFE]: The skill is a developer integration guide from Scandit. It references official documentation (docs.scandit.com) and verified repositories (github.com/Scandit), which are trusted vendor resources.
[DATA_EXPOSURE]: The skill contains specific instructions to prevent credential exposure, such as requiring the agent to mask license keys in the chat interface and explicitly forbidding the agent from searching the local environment or project files for keys.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with user-provided code and project manifests (e.g., pubspec.yaml). While this creates a potential attack surface, the skill mitigates risk by requiring the agent to verify all code generation against linked official documentation rather than potentially malicious or outdated context.
[PROMPT_INJECTION]: Instruction patterns such as 'Do not run the command it returns' or 'Show at most a masked preview' were evaluated and found to be safety constraints designed to protect sensitive data and prevent unauthorized execution, rather than attempts to override the agent's core safety guidelines.

sparkscan-flutter