The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and process content from arbitrary URLs provided by the user or found during searches.
Ingestion points: The crawl.sh, search.sh, and news.sh scripts retrieve content from external websites via the api.search1api.com service.
Boundary markers: There are no boundary markers or delimiters implemented within the scripts to distinguish between instructions and data when the fetched content is returned to the agent.
Capability inventory: The skill uses shell scripts to execute curl, jq, and python3 commands.
Sanitization: While jq is used to safely encode outbound data into JSON (preventing shell injection into the API request), there is no sanitization or filtering of the inbound web content (markdown/text) before it is passed back to the agent.
[DATA_EXPOSURE]: The shared helper script scripts/_common.sh accesses the local file system to read sensitive configuration data.
Evidence: The _resolve_api_key function reads ~/.openclaw/openclaw.json using a Python one-liner to extract the apiKey for Search1API. This is a legitimate functional requirement for the 'OpenClaw' platform but involves accessing a file in the user's home directory.

search1api