The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute a local script, scripts/render.sh, for converting Markdown to HTML. This involves passing user-influenced file paths as arguments to the shell script.
[INDIRECT_PROMPT_INJECTION]: The skill processes content fetched from external URLs and local files, creating a surface for indirect prompt injection attacks.
Ingestion points: Untrusted data is ingested via the WebFetch tool (for URLs) and the Read tool (for local files).
Boundary markers: No specific delimiters or "ignore instructions" markers are used to encapsulate the ingested data within the prompt.
Capability inventory: The skill possesses the ability to fetch network data, write to the filesystem, and execute shell commands through Bash.
Sanitization: The skill lacks explicit sanitization or filtering of the content retrieved from external sources before it is processed by the model.
[EXTERNAL_DOWNLOADS]: The render.sh script invokes npx --yes markmap-cli, which fetches and executes the markmap-cli package from the official NPM registry. This is a well-known service and the package is a standard tool for this use case.
[DYNAMIC_EXECUTION]: The skill performs dynamic execution by using npx to run a CLI tool downloaded at runtime during the optional rendering phase.

mindmap-zh