The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external content such as academic papers, web articles, and reports. This creates a surface where malicious instructions could theoretically be embedded in the source documents to influence the agent's summary or analysis.
Ingestion points: User-provided documents (papers, articles, reports) processed in Workflows 1, 2, and 3 as defined in SKILL.md.
Boundary markers: The skill uses structured templates (IMRAD, claim-evidence-implication), which provide some structure, but it lacks explicit instructions for the agent to ignore hidden commands within the source text.
Capability inventory: The skill can execute local Python scripts (scripts/extract_citations.py, scripts/format_summary.py) and perform file read operations on user-specified paths.
Sanitization: No explicit sanitization or filtering for prompt injection patterns is present in the provided Python scripts.
[COMMAND_EXECUTION]: The skill relies on executing local Python scripts to handle formatting and extraction tasks.
Evidence: SKILL.md contains instructions for the agent to run python3 scripts/format_summary.py and python3 scripts/extract_citations.py to automate the workflow.

research-summarizer