The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill documentation (SKILL.md) provides an installation command that downloads a script from https://cli.search1api.com/install.sh and pipes it directly into the bash interpreter. This pattern is a critical security risk as it executes unverified code from a remote server without manual review.
[COMMAND_EXECUTION]: The skill uses a command-line interface (s1) to execute web searches, crawl pages, and manage user authentication (s1 login). It also suggests the use of npm install -g for tool setup.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill ingests untrusted data from the web via the search, crawl, news, and sitemap functionalities in SKILL.md.
Boundary markers: The instructions advise the agent to cite source URLs and distinguish facts from inferences, but there is no programmatic boundary enforcement during data retrieval.
Capability inventory: The skill includes the s1 learn command, which has file-write capabilities and is used to create and install new skill directories in ~/.agents/skills/.
Sanitization: The agent is instructed to synthesize descriptions and routing tables for these new skills based on the content of the crawled external web pages, creating a path where malicious content could influence generated skill behavior.

search1api