The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill sets up an interactive protocol that ingests user inputs and inserts them into template blocks to generate code snippets, creating a potential surface for prompt or schema malformation.
Ingestion points: User responses to the 11 onboarding questions in SKILL.md (Step 2), such as agent names, capabilities, and budgets.
Boundary markers: Absent. There are no explicit instructions or delimiters telling the agent to escape or isolate user input before inserting it into the code templates.
Capability inventory: The skill doesn't perform automated tool execution; it acts strictly as an interactive code generator and informational reference.
Sanitization: Basic alphanumeric validation is specified for the agent name and Ethereum address format, but other free-form fields lack detailed input sanitization instructions.
[EXTERNAL_DOWNLOADS]: The reference sections contain installation instructions directing users or agents to pull files from external endpoints.
Evidence: Documented commands like curl -sL https://market.agirails.io/skills/claude-code/skill.md and git clone https://github.com/agirails/openclaw-skill.git are suggested for runtime integrations.

agirails-agent-payments