The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs users to install a CLI tool by downloading a shell script from a remote URL and executing it immediately via bash without verification.
Evidence: curl -fsSL https://raw.githubusercontent.com/geekjourneyx/jina-cli/main/scripts/install.sh | bash in SKILL.md.
[EXTERNAL_DOWNLOADS]: The skill requires downloading and compiling code from a third-party GitHub repository (geekjourneyx/jina-cli) that is not owned by a trusted organization or the stated skill author.
Evidence: go install github.com/geekjourneyx/jina-cli/cli@latest in SKILL.md.
[COMMAND_EXECUTION]: The Windows installation instructions include a command to persistently modify the user's PATH environment variable.
Evidence: [Environment]::SetEnvironmentVariable("Path", $env:Path + ";$HOME/go/bin", "User") in SKILL.md.
[PROMPT_INJECTION]: The skill processes untrusted data from the web and social media, creating a significant surface for indirect prompt injection.
Ingestion points: The read and search commands fetch content from arbitrary URLs and AI search results into the agent's context.
Boundary markers: No delimiters or instructions are provided to help the agent distinguish between trusted instructions and untrusted web content.
Capability inventory: The agent can execute the jina CLI, write results to local files using --output-file, and manage persistent configuration keys.
Sanitization: There is no evidence that the fetched content is sanitized or filtered before being processed.

jina-cli