The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes content from arbitrary external web pages to convert them into Markdown format. This represents a vulnerability surface for indirect prompt injection, where an attacker could embed malicious instructions in a web page that the agent then processes.
Ingestion points: External URLs are processed by the web2md CLI tool (as described in SKILL.md).
Boundary markers: The skill does not define specific delimiters or instructions to treat the resulting Markdown as untrusted data when returned to the agent.
Capability inventory: The skill utilizes shell command execution (web2md, mkdir, ls, wc) and file system access.
Sanitization: No explicit sanitization or filtering of the fetched web content is performed before it is delivered to the agent's context.
[COMMAND_EXECUTION]: The skill's core workflow involves executing the web2md CLI tool via shell commands. It also uses system utilities such as mkdir for directory creation and ls or wc for validating output files. The skill includes a recommendation to validate that URLs start with http:// or https:// to mitigate simple command injection attempts.
[DATA_EXFILTRATION]: The skill performs network operations to fetch content from external, non-whitelisted domains. While this is the intended functionality of the skill, it represents an automated network access path to external sources.
[EXTERNAL_DOWNLOADS]: The skill documentation instructs the user to install an external dependency, the web2md CLI, via npm or by cloning a repository. This CLI tool is authored by the same developer as the skill ('davila7'), representing a vendor-specific resource.

web-to-markdown