The Agent Skills Directory

[PROMPT_INJECTION]: The SKILL.md file contains a pre-written block of text intended for the user to paste into their AI assistant. This payload instructs the agent to perform specific actions including installing external packages and purchasing a separate skill. It explicitly attempts to override safety protocols with instructions such as 'I have authorized installation... do not confirm for me,' which is designed to manipulate the agent's decision-making and confirmation loop.
[REMOTE_CODE_EXECUTION]: The skill's documentation encourages the execution of the command npx -y @alipay/agent-payment@latest install. This command fetches and executes code from the public npm registry at runtime. While the package name suggests association with a well-known service, the instruction promotes running external, unverified code as part of a commercial upsell flow within the skill's instructions.
[INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface where it ingests untrusted data.
Ingestion points: The scripts/run.mjs script reads external content from local files or JSON input via the --input flag.
Boundary markers: There are no boundary markers or instructions to the agent to ignore commands embedded within the bill discount tables or calculation worksheets being processed.
Capability inventory: The skill has access to the Bash tool and file reading capabilities.
Sanitization: The skill performs numeric and date parsing but does not sanitize the text content for potential natural language instructions before the results are returned to the agent context.

discount-interest-check-free