The Agent Skills Directory

[OBFUSCATION]: The title of the skill in SKILL.md contains a sequence of zero-width Unicode characters (including U+200B, U+200C, U+200D, and U+2060). These characters are used to encode data (likely ASCII bits) that is hidden from the user interface but can be interpreted by the underlying language model. This technique is a characteristic marker for steganographic prompt injection or unauthorized tracking.
Evidence: Found in the string # Mark Minervini Swing Trading Style Guide... in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external stock market data provided in pd.DataFrame format. While the provided Python code performs numerical technical analysis, the consumption of external, untrusted data into the agent's context presents a vulnerability surface for indirect prompt injection if that data contains malicious natural language instructions that the model might later execute.
Ingestion points: The TrendTemplateScanner and VCPDetector classes in SKILL.md take df: pd.DataFrame as input for analysis.
Boundary markers: No delimiters or instructions are present to prevent the agent from following potential commands embedded within the data fields.
Capability inventory: The skill includes functions for technical scanning and risk management calculations but does not demonstrate immediate network exfiltration or file system manipulation capabilities in the provided snippet.
Sanitization: There is no sanitization logic to verify the integrity or safety of the content within the input dataframes.

minervini-swing-trading