The Agent Skills Directory

REMOTE_CODE_EXECUTION (HIGH): The skill provides instructions to run curl -fsSL https://cli.inference.sh | sh. Piping a remote script directly into a shell is a high-risk pattern that allows arbitrary code execution from a source not listed in the trusted repository list.
COMMAND_EXECUTION (HIGH): The allowed-tools section grants the skill permission to run Bash(infsh *), which allows the execution of any CLI command associated with the infsh tool, including logging in and running arbitrary remote scripts via the python-executor app.
EXTERNAL_DOWNLOADS (MEDIUM): The skill uses npx skills add to fetch and install external skills from inferencesh/skills@web-search. This introduces third-party dependencies that are not pre-verified.
DATA_EXFILTRATION (LOW): The skill facilitates the transmission of search queries and potentially user-provided context to external APIs such as Tavily and Exa. While this is part of its functional design, it establishes a data flow to third-party services.
INDIRECT_PROMPT_INJECTION (LOW): The skill has a vulnerability surface for indirect prompt injection by ingesting untrusted web data.
Ingestion points: Search results from tavily/search-assistant and exa/search are incorporated into the agent's context in SKILL.md.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the research tool calls.
Capability inventory: The skill has broad capabilities to execute code via infsh app run infsh/python-executor.
Sanitization: No sanitization of search results is performed before the data is processed by the agent.

case-study-writing