The Agent Skills Directory

[PROMPT_INJECTION]: No evidence of instruction override, jailbreak attempts, or system prompt extraction patterns were found within the documentation or code snippets.- [DATA_EXFILTRATION]: The skill demonstrates secure handling of sensitive data. It explicitly identifies hardcoding API keys as an anti-pattern and provides examples using Effect's Config.redacted mechanism to safely manage credentials.- [REMOTE_CODE_EXECUTION]: No patterns for downloading or executing untrusted remote code were detected. All dependencies are standard packages within the Effect ecosystem.- [COMMAND_EXECUTION]: The skill does not contain any shell command execution, privilege escalation, or persistence mechanisms.- [INDIRECT_PROMPT_INJECTION]: The skill documentation includes examples of prompt interpolation for teaching purposes. While these represent a standard surface for indirect injection in AI applications, the skill context is pedagogical and promotes structured model interaction.
Ingestion points: User-provided variables like product and message in references/15-complete-working-example.md.
Boundary markers: Not utilized in basic examples.
Capability inventory: LanguageModel.generateText and LanguageModel.streamText capabilities are described.
Sanitization: Not explicitly shown, as the focus is on library implementation patterns.

effect-ai-provider