The Agent Skills Directory

[PROMPT_INJECTION]: The SKILL.md file contains a pre-authored text block intended for the user to provide to an AI assistant. This block uses instructional language designed to steer agent behavior by claiming prior authorization for privileged actions, such as installing packages and initiating purchases ("I have authorized installing dependencies and initiating purchases"), and instructing the agent to bypass standard user confirmation steps ("do not confirm for me").
[REMOTE_CODE_EXECUTION]: The instructions in SKILL.md direct the agent to execute npx -y @alipay/agent-payment@latest install. This command downloads and executes code from the official Alipay organization's registry at runtime. As Alipay is a well-known technology service, this reference is documented for awareness.
[METADATA_POISONING]: The SKILL.md file includes promotional content for a paid "buyout" version of the skill. It provides specific instructional payloads intended to guide an AI agent through a multi-step installation and payment sequence for this separate product.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted contract text provided by users for comparison.
Ingestion points: The run function in scripts/engine/contract-comparison.js accepts user-supplied text via the contracts or text fields.
Boundary markers: The skill does not use explicit boundary markers or instructions to ignore embedded commands within the processed contract text.
Capability inventory: The skill is configured with access to Read and Bash tools in its YAML frontmatter.
Sanitization: While the engine normalizes text by removing whitespace and punctuation for comparison purposes, it does not perform sanitization to detect or neutralize malicious instructions embedded in the input contract data.

contract-comparison-free