The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides setup instructions that involve downloading shell and PowerShell scripts from inflowcli.ai and executing them directly via piped shell commands (e.g., curl | bash and irm | iex). While these originate from the vendor's domain, they represent high-privilege remote code execution patterns.
[COMMAND_EXECUTION]: The skill relies on the inflow binary and the brew package manager to perform its core functions and management tasks.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to fetch installation scripts, binary components, and service metadata from inflowcli.ai and other external service origins encountered during discovery.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and processes content from third-party services. This risk is acknowledged in the 'Security and data handling' section of the skill.
Ingestion points: Remote service descriptions, product attributes, schemas, and action definitions processed by the inflow odp commands (e.g., SKILL.md).
Boundary markers: The skill does not define technical delimiters or specific instruction-ignore patterns for the interpolated data.
Capability inventory: The agent has access to shell tools including inflow, brew, and curl to interact with external data.
Sanitization: No programmatic sanitization or validation of the remote service output is described; the skill relies on text-based warnings to the agent to treat content as untrusted.

agentic-discovery