The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The SKILL.md file contains a command to download and execute a shell script from https://cli.inference.sh. This is an installation instruction for the infsh CLI. The skill's environment is restricted to the infsh tool, which prevents the agent from executing the curl | sh pattern itself during runtime.
[EXTERNAL_DOWNLOADS]: The skill references external binaries from dist.inference.sh and additional skills from the inference-sh registry via npx skills add. These represent standard dependencies within the vendor's ecosystem and are documented for user setup.
[COMMAND_EXECUTION]: The skill uses the Bash tool to run infsh app commands. Execution is limited to the infsh binary via the allowed-tools frontmatter field, which is a security best practice for restricting shell access.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a proxy for external tools that ingest market research data and user-defined prompts.
Ingestion points: Research queries (Step 1) and avatar descriptions (Step 8) in SKILL.md are derived from user requirements.
Boundary markers: None explicitly defined in the instruction text; data is passed as JSON inputs to the infsh tool.
Capability inventory: Web searching via Tavily and Exa, and image generation via Fal.ai.
Sanitization: The skill relies on downstream AI services and the CLI tool itself to sanitize research results and generated image content.

customer-persona