The Agent Skills Directory

[SAFE]: The skill provides logic for academic reference verification and BibTeX generation. The analysis found no evidence of prompt injection, data exfiltration, or unauthorized command execution. The skill operates within a local workspace and does not perform network operations.
[INDIRECT_PROMPT_INJECTION]: The skill ingests content from paper drafts, external metadata, and PDF files, which constitutes an attack surface for indirect prompt injection. However, the risk is minimal as the skill only performs text processing and file writing without executing the ingested content.
Ingestion points: The skill reads files from paper/sections/, workspace/papers/related_paper_analysis.md, and PDF files in workspace/papers/ as defined in the Inputs and Workflow sections.
Boundary markers: There are no explicit delimiters or safety instructions provided to the agent to distinguish between its instructions and the processed data.
Capability inventory: Capabilities are restricted to reading local files and writing the refs.bib and reference_audit.md files. No network tools or subprocess execution tools are requested or used.
Sanitization: No input validation or output sanitization is specified for the data extracted from source files before it is included in the output documents.

reference-manager