The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands incorporating user-supplied or externally-sourced URLs verbatim within single quotes (e.g., web2md '<url>'). The instructions lack requirements for escaping single quotes or other shell-sensitive characters that may be present in a malicious URL. The required validation check (verifying the URL starts with http:// or https://) is insufficient to prevent injection attacks (e.g., https://example.com' ; malicious_command ; ').
[INDIRECT_PROMPT_INJECTION]: The skill fetches and processes content from arbitrary external web pages, creating a surface for instructions embedded in web content to influence the agent.
Ingestion points: Web content is fetched via Puppeteer and converted to Markdown (SKILL.md, Workflow step 4).
Boundary markers: None identified; the instructions do not provide delimiters or "ignore instructions" warnings for the agent when it subsequently processes the generated Markdown.
Capability inventory: The agent can execute the web2md shell command and perform file system writes via the --out parameter (SKILL.md, Workflow step 4).
Sanitization: The instructions only mandate a protocol prefix check and do not specify sanitization of the fetched web content or escaping of the URL before it is passed to the shell.
[EXTERNAL_DOWNLOADS]: The skill describes installation procedures for a local CLI tool using npm install and npm link within a local project directory (~/workspace/softaworks/projects/web2md). These operations utilize the standard NPM registry for dependencies.

web-to-markdown