The Agent Skills Directory

[PROMPT_INJECTION]: The instructions do not contain any commands to bypass safety filters, override system prompts, or disregard previous instructions. The language is purely functional and focused on the stated task of code generation.
[DATA_EXFILTRATION]: No network activity was detected. The skill does not contain any curl, wget, or similar commands. It manages user activity data locally using standard UserDefaults keys for the generated Swift application.
[REMOTE_CODE_EXECUTION]: The skill does not download external scripts or packages. It relies entirely on provided Markdown templates to generate local Swift files.
[COMMAND_EXECUTION]: While the skill requests Bash and Write capabilities, these are scoped to project context discovery (e.g., searching for existing Swift files) and writing the generated feature code to the local filesystem. No dangerous or privileged commands were found.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes existing project files via Glob and Grep to determine how to integrate the generated code. While this involves reading untrusted local data, the search patterns are specific to development artifacts (like lastActiveDate) and do not process data in a way that would trigger unintended agent actions.
[OBFUSCATION]: The skill content is provided in plain text. No Base64, hex encoding, zero-width characters, or homoglyphs were used to hide instructions or URLs.

lapsed-user