The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/md_to_pdf.py utilizes the subprocess.run function to execute local browser binaries (such as Google Chrome, Chromium, or Microsoft Edge) to perform PDF rendering tasks.
[COMMAND_EXECUTION]: The browser execution command includes the --no-sandbox flag. This flag disables the primary security sandbox of the browser, which significantly increases the risk of host system compromise if the content being rendered contains malicious exploits.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface by processing untrusted data from the WeChat Reading API without sanitization or clear boundaries.
Ingestion points: Untrusted data is retrieved from the /book/bookmarklist and /review/list/mine API endpoints.
Boundary markers: The skill lacks delimiters or specific instructions to the agent to treat ingested book content as non-instructional data.
Capability inventory: The skill has capabilities for file system writes via Path.write_text and shell command execution via the PDF conversion script.
Sanitization: No HTML escaping or text validation is performed on the ingested highlights and thoughts before they are rendered within the headless browser context.

read-weread-export