The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes Python scripts to manage local file operations and execute academic search tasks. The scripts/common.py file implements robust path validation functions (is_within, require_within) that are used across all scripts to ensure file writes and reads remain confined to the authorized workspace directory, effectively preventing path traversal attacks.
[EXTERNAL_DOWNLOADS]: The scripts/fetch_pubmed_recent.py script performs network requests to the official NCBI E-Utilities API (eutils.ncbi.nlm.nih.gov). These requests are used solely to fetch metadata for recent academic articles (titles, abstracts, PMIDs). The implementation uses standard Python libraries and targets a well-known scientific service, which is considered a safe and intended functionality for this skill.
[DYNAMIC_EXECUTION]: A dynamic import of the standard datetime library was identified in scripts/shortlist_journals.py (__import__("datetime")). This is a benign coding pattern used for inline timestamp generation and does not facilitate unauthorized code execution as the imported module is a hardcoded standard library.
[INDIRECT_PROMPT_INJECTION]: The skill represents an ingestion surface for untrusted data because it processes user-provided manuscripts (abstracts, full text, or files). The ingested content is used to inform AI-driven recommendations. While this is a common vector for indirect prompt injection, the skill limits its capabilities to academic reporting and data retrieval from trusted sources, and the scripts utilize structured JSON processing which minimizes the risk of control-flow manipulation.
[DATA_EXPOSURE_&_EXFILTRATION]: No sensitive local file paths (e.g., .ssh, .aws, .env) are accessed. Network communication is restricted to the legitimate PubMed API for data retrieval related to the skill's primary purpose. No hardcoded credentials or exfiltration patterns were found.

paper-select-journal