The Agent Skills Directory

REMOTE_CODE_EXECUTION (CRITICAL): The skill documentation includes a command to download and execute a script directly from https://cli.inference.sh using curl -fsSL ... | sh. This is a high-risk pattern that executes remote code with the user's local privileges without any integrity verification.
EXTERNAL_DOWNLOADS (HIGH): The skill promotes installing additional dependencies using npx skills add from the inference-sh organization. As this organization is not on the trusted sources list, these downloads are considered unverified and potentially malicious.
COMMAND_EXECUTION (MEDIUM): The skill is configured with allowed-tools: Bash(infsh *), which permits the execution of any subcommand under the infsh umbrella. This provides a significant attack surface for local command execution via the agent.
PROMPT_INJECTION (LOW): The skill's workflow involves ingesting untrusted data from external search providers (Tavily, Exa).
Ingestion points: Research data is pulled into the agent context via infsh app run commands.
Boundary markers: None identified in the prompt templates to distinguish between instructions and search data.
Capability inventory: Full bash access via infsh and the ability to write files (implied by press release output).
Sanitization: No sanitization logic is present to filter malicious instructions embedded in search results.

press-release-writing