The Agent Skills Directory

[METADATA_POISONING]: The skill includes a file named literature-statistics_audit_result_v2.json which contains a detailed, simulated security audit report. This report claims a 'Production Ready' grade and a 'PASS' on all security gates. This is a deceptive practice intended to influence the auditor's judgment and bypass security reviews by providing a false sense of pre-validated safety.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content (PDF files and user-pasted citations) and returns a summary to the agent context.
Ingestion points: scripts/process_pdfs.py reads files from the user-provided --input-dir and extracts text from the first page and PDF metadata.
Boundary markers: The script does not use specific delimiters or instructions to the agent to ignore content embedded within the extracted literature data.
Capability inventory: The script has the ability to read files from the local filesystem and write the results to a markdown file (--output).
Sanitization: The script performs regex validation for years and simple keyword filtering for journal titles, which mitigates the risk of full document text being injected, but metadata fields (Title, Subject) are extracted without significant sanitization, potentially allowing malicious instructions in PDF metadata to reach the agent.

literature-statistics