The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill documents the use of the pickle module (pickle.load, nx.read_gpickle) in references/io.md. Using pickle to deserialize data from untrusted sources is unsafe as it allows for arbitrary code execution.\n- [PROMPT_INJECTION]: Potential for indirect prompt injection via external data files.\n
Ingestion points: Functions such as nx.read_edgelist, nx.read_graphml, and nx.read_json in references/io.md.\n
Boundary markers: Absent; no specific delimiters or warnings to ignore embedded instructions are provided to the agent.\n
Capability inventory: The skill enables file system writes (nx.write_edgelist), database modifications (df.to_sql), and interactive visualizations via Plotly or PyVis.\n
Sanitization: Absent; the documentation does not describe methods for escaping or validating attributes read from external files.\n- [DATA_EXFILTRATION]: Extensive file and database I/O capabilities documented in references/io.md and SKILL.md (e.g., nx.read_edgelist, nx.write_graphml, pd.to_sql) allow for accessing and writing sensitive local data.

networkx