The Agent Skills Directory

[Prompt Injection] (SAFE): No instructions were found that attempt to override AI safety constraints, reveal system prompts, or bypass ethical protocols.
[Data Exposure & Exfiltration] (SAFE): The skill does not access sensitive file paths (e.g., credentials or SSH keys) and contains no network-bound code.
[Obfuscation] (SAFE): All instructions are in plain markdown; no Base64, zero-width characters, or homoglyphs were detected.
[Unverifiable Dependencies & Remote Code Execution] (SAFE): No package managers are invoked, and no remote scripts are downloaded or executed.
[Privilege Escalation] (SAFE): No commands like sudo or chmod are present.
[Persistence Mechanisms] (SAFE): No attempts to modify shell profiles, cron jobs, or registry keys.
[Metadata Poisoning] (SAFE): The YAML metadata is standard and contains no hidden instructions.
[Indirect Prompt Injection] (SAFE): 1. Ingestion points: The skill processes user-supplied post text. 2. Boundary markers: Absent. 3. Capability inventory: The skill is limited to text generation (display-only) and has no access to tools or command execution. 4. Sanitization: Absent. The risk is negligible because the skill lacks the capabilities to act on injected instructions.
[Time-Delayed / Conditional Attacks] (SAFE): No logic exists to trigger behavior based on dates or environmental conditions.
[Dynamic Execution] (SAFE): No runtime code generation, compilation, or unsafe deserialization occurs.

making-waffles