The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes relational datasets containing node and edge attributes, which represents a potential surface for indirect prompt injection if those attributes contain malicious instructions that the agent processes or reports.
Ingestion points: External network data (relational ties and attributes) as described in SKILL.md.
Boundary markers: The skill does not define specific delimiters or instructions to ignore content within network attribute fields.
Capability inventory: Access to local Python execution via uv and R execution via a bridge system as specified in SKILL.md.
Sanitization: There are no explicit sanitization routines defined for node/edge metadata before processing.
[DYNAMIC_EXECUTION]: The skill facilitates the generation and execution of R scripts via an external bridge to perform Exponential Random Graph Modeling (ERGM), as native Python support for these models is noted as immature.
Evidence: SKILL.md and references/sna_methods.md provide R code templates for statnet/ergm and instruct the agent to use an R bridge for computation.
[COMMAND_EXECUTION]: Analysis workflows are executed locally using the uv package manager to run Python scripts.
Evidence: The compatibility section in SKILL.md explicitly mentions running locally via uv run python.

alterlab-sna