The Agent Skills Directory

[PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection. It is designed to ingest and process data from arbitrary external websites. * Ingestion points: scripts/recon-site.mjs, scripts/route-crawl.mjs, and scripts/interaction-probe.mjs are used to load external URLs. * Boundary markers: None. The instructions do not provide delimiters or guidance for the agent to ignore instructions embedded in the target site's content. * Capability inventory: The skill allows for file creation, shell command execution, and network access. * Sanitization: None. The skill does not include mechanisms to sanitize or filter the content retrieved from target sites.
[EXTERNAL_DOWNLOADS]: The skill involves downloading source code and a wide range of assets from remote servers via scripts/asset-harvest.mjs and scripts/mirror-site.mjs.
[REMOTE_CODE_EXECUTION]: The skill workflow involves downloading remote source code and explicitly suggests that the user execute it locally to verify results. Furthermore, scripts/lib/playwright-loader.mjs attempts to load dependencies from several hardcoded absolute file paths on the local system (e.g., /Users/jane/ZenDrop/node_modules/playwright), which represents an insecure dynamic loading pattern.
[COMMAND_EXECUTION]: The skill requires the agent to execute numerous system commands, such as git clone, node, curl, and gh, to initialize projects and run reconnaissance tools.

web-clone