The Agent Skills Directory

[SAFE]: The skill is entirely defensive, providing high-quality tutorial implementations for token bucket, sliding window, and adaptive rate limiters along with an NGINX configuration guide. All code samples are static and safe for implementation review.
[INDIRECT_PROMPT_INJECTION]: Risk analysis of external log ingestion:
Ingestion points: The log analysis script scripts/agent.py reads user-provided log files through load_access_logs.
Boundary markers: Absent from the log parser as it reads arbitrary line-delimited JSON.
Capability inventory: There are no dangerous tools, subprocess calls, database executions, or network connections available in the script.
Sanitization: The input data is handled strictly as structural strings via json.loads and stored into aggregate statistics, preventing any code execution or command injection surfaces.

implementing-api-abuse-detection-with-rate-limiting