The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions to download and install external binaries from the vendor's repository. Specifically, it recommends using npx -y @mvanhorn/printing-press-library install exa --cli-only or go install github.com/mvanhorn/printing-press-library/library/ai/exa/cmd/exa-pp-cli@latest. These resources are managed by the skill's authoring organization.
[DATA_EXFILTRATION]: The CLI tool includes a --deliver webhook:<url> flag that allows command output to be POSTed to an external URL. This feature provides a direct mechanism for data exfiltration of search results or local database content if the agent is directed to use it maliciously. Additionally, the tool allows setting a custom EXA_FEEDBACK_ENDPOINT for sending feedback data.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from the live web (e.g., websearch, contents, answer).
Ingestion points: The websearch, contents, and answer commands retrieve external content from the Exa API into the agent's context.
Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands within the retrieved search results.
Capability inventory: The skill has the ability to write to the local filesystem (--deliver file:<path>) and make network requests (--deliver webhook:<url>).
Sanitization: There is no mention of sanitization or filtering of the external web content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill requires the Read Bash tool to execute the exa-pp-cli binary and perform system-level operations such as environment variable management and local database interactions. The instructions include commands that modify the environment (e.g., EXA_HOME, EXA_CONFIG_DIR).

pp-exa