The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from user-provided arguments and external sources (web searches and PDF documents) which is then used to drive the behavior of Author, Editor, and Reader agents.
Ingestion points: The skill utilizes $ARGUMENTS for initial user input and performs external research via web searches and PDF data extraction in Phase 1.
Boundary markers: The agent prompts do not implement explicit delimiters or 'ignore embedded instructions' warnings for the external content being processed.
Capability inventory: The skill has the capability to execute shell commands (pdftoppm, date), invoke local Python scripts (tools/financial_rigor.py), and write files to the local file system (reports/, assets/).
Sanitization: No specific sanitization, filtering, or validation logic is specified for the external data before it is interpolated into sub-agent prompts.
[COMMAND_EXECUTION]: The skill instructions involve the execution of local system utilities to support its functional requirements.
Evidence: The skill runs the date command to establish a research baseline and uses pdftoppm to extract high-resolution images from academic papers for inclusion in the generated articles.

wechat-article