The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the @upstash/cli package globally via NPM. This is a standard installation of the official tool provided by the vendor (Upstash) and does not involve untrusted third-party sources.
[COMMAND_EXECUTION]: The skill consists primarily of shell command templates for the upstash CLI and AWS CLI. These commands perform legitimate administrative tasks such as creating databases, managing team members, and interacting with object storage. The skill also recommends using --dry-run for destructive actions like deletions.
[CREDENTIALS_UNSAFE]: The documentation mentions standard locations for secret storage, such as ~/.config/upstash/config.json and .env files. These references are part of the tool's documented authentication flow and do not indicate malicious harvesting or exfiltration of credentials. It also correctly advises the use of read-only keys for agents where possible.
[INDIRECT_PROMPT_INJECTION]: The skill encourages piping CLI output to jq for data extraction. While processing external tool output is an inherent attack surface for LLMs, the skill does not contain any specific patterns designed to exploit this surface or bypass safety filters.

upstash-cli