The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes instructions to bypass a specific platform execution model (the seven-step Algorithm workflow) in favor of a 'NATIVE' mode. While this overrides standard operating procedures, it is done to streamline the specific multi-tool pipeline rather than to bypass safety or ethical guidelines.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests untrusted data from external sources, including ArXiv URLs, paper PDFs, and web links.
Ingestion points: User-provided paper links and file paths are extracted and passed to subsequent tools.
Boundary markers: The instructions do not define explicit delimiters or 'ignore' instructions for the content of the processed papers.
Capability inventory: The skill has the capability to execute the ljg-paper and ljg-card tools (SKILL.md).
Sanitization: There is no mention of sanitization or validation of the text extracted from the external papers before it is processed by the subagents.
[COMMAND_EXECUTION]: The skill orchestrates the execution of two vendor-prefixed tools, ljg-paper and ljg-card. This usage is consistent with the author's identity ('lijigang') and represents the core intended functionality of a workflow aggregator.

ljg-paper-flow