The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill documents the use of pickle.load() and nx.read_gpickle() for serializing and deserializing graph objects in references/io.md. Deserialization with the pickle module is an unsafe operation that can lead to arbitrary code execution if the input files originate from an untrusted source.
[DATA_EXFILTRATION]: The skill provides numerous methods for reading from and writing to the local file system and SQL databases, as documented in references/io.md. These capabilities allow the agent to access and move potentially sensitive data between different files, databases, and its memory.
[PROMPT_INJECTION]: The skill features a broad interface for ingesting external data (CSV, JSON, GraphML, SQL, and others), which introduces a surface for indirect prompt injection.
Ingestion points: Numerous functions in references/io.md such as read_edgelist, read_graphml, and from_pandas_edgelist.
Boundary markers: Absent from the documented examples and instructions.
Capability inventory: File system writes, SQL database interaction, and execution of visualization tools (matplotlib, plotly).
Sanitization: No sanitization or validation of input data is required or demonstrated.
[EXTERNAL_DOWNLOADS]: The skill suggests the installation of the networkx library and its optional dependencies (such as matplotlib, scipy, and pandas) from standard package registries.

networkx