The Agent Skills Directory

[Prompt Injection] (SAFE): No instructions were found that attempt to bypass safety guidelines, override system behavior, or extract system prompts. The instructions are purely professional and context-specific for system design.
[Data Exposure & Exfiltration] (SAFE): The skill does not reference sensitive file paths, include hardcoded credentials, or perform any network operations. It is a documentation-based guidance skill.
[Obfuscation] (SAFE): No encoded strings, homoglyphs, zero-width characters, or other methods of hiding malicious content were detected.
[Unverifiable Dependencies & Remote Code Execution] (SAFE): There are no commands for installing third-party packages (pip, npm) or downloading/executing remote scripts.
[Privilege Escalation] (SAFE): The skill does not use administrative commands such as sudo or attempt to modify system permissions.
[Persistence Mechanisms] (SAFE): No attempts to modify shell profiles, cron jobs, or startup services were found.
[Metadata Poisoning] (SAFE): Metadata fields like name and description accurately reflect the skill's purpose and do not contain hidden malicious instructions.
[Indirect Prompt Injection] (SAFE): While the skill discusses RAG (which involves processing external data), the skill itself does not ingest or process untrusted data; it only provides design templates and constraints for the architect role.
[Time-Delayed / Conditional Attacks] (SAFE): No logic exists within the skill that triggers specific behaviors based on date, time, or environment variables.
[Dynamic Execution] (SAFE): The skill contains no code-generation logic, runtime compilation, or unsafe data deserialization patterns.

rag-architect