The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install standard, well-known packages @upstash/ratelimit and @upstash/redis via npm. These are official client libraries for a well-known service provider.
[CREDENTIALS_UNSAFE]: The skill correctly advises using environment variables (UPSTASH_REDIS_REST_URL and UPSTASH_REDIS_REST_TOKEN) for sensitive configuration rather than hardcoding them, which follows security best practices.
[INDIRECT_PROMPT_INJECTION]: The skill describes an attack surface where rate limiting is applied to identifiers derived from untrusted data (like the x-forwarded-for header).
Ingestion points: The middleware example (SKILL.md) reads the x-forwarded-for header from incoming requests.
Boundary markers: The skill explicitly warns that clients can spoof headers like x-forwarded-for and that they must be validated.
Capability inventory: The skill interacts with Upstash Redis via a REST client to update counters and check limits.
Sanitization: The skill highlights the need for header validation in its "Best Practices" section.
[DATA_EXFILTRATION]: While the skill performs network operations to Upstash Redis, these are the primary intended purpose of the skill (distributed rate limiting). The data sent consists of application-defined identifiers (like user IDs or IP addresses) and does not involve exfiltration of system credentials or sensitive files.

upstash-ratelimit