The Agent Skills Directory

[SAFE]: The skill facilitates searches across 18 well-known and reputable scholarly APIs including PubMed, arXiv, Crossref, and OpenAlex. All network activity is directed toward established academic and governmental domains.
[REMOTE_CODE_EXECUTION]: Automated security alerts regarding remote code execution are false positives. The flagged commands (e.g., curl ... | python3 scripts/arxiv_atom.py -) use local, bundled Python scripts to parse XML or JSON data retrieved from well-known scholarly sources. These scripts function as parsers and do not execute the content of the downloaded data.
[CREDENTIALS_UNSAFE]: The skill employs proactive credential protection. The shared utility scripts/_common.py and the pagination script scripts/paginate.py both include logic to detect and redact API keys and sensitive parameters (such as email, mailto, and api_key) from any provenance information emitted to the agent or user.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external scholarly APIs (ingestion points in scripts/paginate.py and curl examples). It includes a mandatory instruction in SKILL.md for the agent to treat all responses as untrusted and to ignore any instructions found within the data. Additionally, scripts/_common.py provides sanitization by stripping control characters from external text.
[DYNAMIC_EXECUTION]: The provided Python scripts use only the standard library and follow safe parsing practices for JSON and XML data, avoiding dangerous functions like eval, exec, or unsafe deserialization.

paper-lookup