The Agent Skills Directory

[PROMPT_INJECTION]: The SKILL.md file contains instructions for the AI agent that use steering language, specifically asserting that the user has already "authorized installing dependencies and initiating purchase." This technique is used to influence the agent's behavior to execute shell commands for software installation with reduced user scrutiny.
[REMOTE_CODE_EXECUTION]: Documentation within the skill encourages the use of npx -y @alipay/agent-payment@latest install, which downloads and executes code from the NPM registry. The package is managed by a well-known financial technology organization.
[EXTERNAL_DOWNLOADS]: The skill references external URLs, including skillpay.alipay.com for service onboarding and chenqg618.github.io for author-provided documentation. These resources are associated with a well-known service and the skill's author, respectively.
[COMMAND_EXECUTION]: The skill provides instructions for running local scripts via node scripts/run.mjs to perform its primary function of tax rebate calculation.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data provided by users via the --input argument, which creates a potential surface for indirect injection attacks. 1. Ingestion points: Data is read from files or strings passed to the run.mjs script via user input. 2. Boundary markers: The system does not utilize specific delimiters or instructions to the agent to ignore potentially malicious content within the tax data table. 3. Capability inventory: The skill has access to shell execution (Bash) and file system access (Read) across its core scripts. 4. Sanitization: The engine performs numeric parsing and regex-based table extraction, which acts as a basic filter, but does not provide comprehensive sanitization against instruction-based attacks.

export-rebate-check-free