The Agent Skills Directory

[SAFE]: The skill is a specialized tool for academic citation analysis. It focuses on citation formatting and metadata accuracy and does not perform any suspicious background activities.
[COMMAND_EXECUTION]: The skill utilizes a bundled Python script, scripts/citation_audit.py, to perform automated parsing of citations and reference lists. This script processes document text locally and does not exhibit dangerous capabilities like network access or unauthorized file modification.
[EXTERNAL_DOWNLOADS]: The skill references well-known academic metadata services including Crossref, ORCID, and the ICMJE recommendations. These references are used for verification of scholarly metadata and are from established, well-known providers.
[SAFE]: Analysis of the indirect prompt injection surface identifies that the skill processes untrusted user-provided manuscripts. This is handled as follows:
Ingestion points: User-provided manuscript files and related research documents mentioned in SKILL.md.
Boundary markers: No specific delimiters are used to wrap user content.
Capability inventory: Local execution of the bundled parsing script citation_audit.py.
Sanitization: Input content is parsed for metadata patterns using regular expressions; no content is treated as executable code.

rw-citation-audit