The Agent Skills Directory

[DATA_EXFILTRATION]: The skill performs network operations to external websites using tools like agent-browser and curl. The documentation includes specific references to an external domain (https://www.angelokeirsebilck.be/) in examples demonstrating how to handle external navigation triggers.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest, record, and process content from arbitrary third-party websites, which serves as a potential attack surface if the target site contains adversarial content.
Ingestion points: Website pages and DOM snapshots accessed via the agent-browser tool during the exploration phase (SKILL.md).
Boundary markers: The skill implements a mandatory 'Authorization Gate' requiring user confirmation of rights before processing, and includes instructions to avoid login-protected areas (SKILL.md).
Capability inventory: The skill utilizes bash for directory creation and curl for downloading external assets, alongside agent-browser for navigation and screenshot capture (SKILL.md).
Sanitization: The instructions do not define specific mechanisms for sanitizing or escaping text and attributes captured from the DOM before they are integrated into the final replication blueprint or code (SKILL.md).
[COMMAND_EXECUTION]: The skill instructions utilize the bash tool to perform local filesystem operations, including directory creation (mkdir) and asset retrieval via curl (SKILL.md).

glmv-web-replication