The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill enables the agent to fetch and process untrusted content from the web, creating a surface for indirect prompt injection.
Ingestion points: Search results retrieved from the Tavily API in SKILL.md.
Boundary markers: The skill does not explicitly define delimiters or instructions for the agent to ignore embedded commands within search results.
Capability inventory: Includes access to the local shell for tool management (hermes tools enable) and network operations via curl.
Sanitization: There are no documented steps for sanitizing or filtering the content returned by the external API.
[COMMAND_EXECUTION]: The documentation provides copy-paste shell commands for the user to verify their API configuration.
Evidence: SKILL.md includes a curl command that reads a local file to send an authenticated request to api.tavily.com.
[DYNAMIC_EXECUTION]: The skill uses a Python one-liner to dynamically parse local JSON configuration files at the command line.
Evidence: python3 -c "import json; print(json.load(open('$HOME/.config/tavily/credentials.json'))['api_key'])" is used to extract the API key from a file.

web-search