The Agent Skills Directory

[PROMPT_INJECTION] (LOW): Susceptible to indirect prompt injection where malicious instructions embedded in academic abstracts could influence the LLM's output. Ingestion points: Semantic Scholar and arXiv search results. Boundary markers: None specified. Capability inventory: Execution of local Python scripts and file system writes for paper generation. Sanitization: No evidence of input filtering or sanitization of external metadata.\n- [COMMAND_EXECUTION] (SAFE): Executes a local Python script search_semantic_scholar.py. The script path is relative to the user's skill directory and is used to facilitate the primary academic search function.\n- [EXTERNAL_DOWNLOADS] (SAFE): Accesses Semantic Scholar and arXiv APIs to retrieve research data. These are trusted academic sources and the skill does not execute code downloaded from these locations.

survey-generation