The Agent Skills Directory

[SAFE]: The skill defines a standard set of operations for managing graph data through a local API endpoint (http://localhost:3411). No evidence of credential theft, obfuscation, or remote code execution was found.
[PROMPT_INJECTION]: The skill was evaluated for potential indirect prompt injection as it processes external graph data.
Ingestion points: API responses from get_graph_description and get_graph_node in SKILL.md are ingested into the agent context.
Boundary markers: The instructions do not specify explicit delimiters for untrusted data, though they emphasize schema-based validation.
Capability inventory: The skill possesses the capability to modify graph states via apply_graph_transaction in SKILL.md.
Sanitization: The skill mandates validation against registered node type schemas before submitting transactions.
Analysis: While a data ingestion surface exists, the risk is inherent to the skill's purpose and is mitigated by the local nature of the API and structural validation requirements.

graph-management