The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes curl for interacting with LiteLLM proxy endpoints and python3 for parsing and formatting JSON results. These tools are used appropriately for the skill's primary function of usage reporting and are restricted via the allowed-tools frontmatter.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from a remote LiteLLM proxy, creating a potential surface for indirect prompt injection if the source data is compromised.
Ingestion points: Data is ingested via curl requests to the user-configured LiteLLM proxy URL in SKILL.md.
Boundary markers: Absent; the skill does not use specific markers to delimit external data from agent instructions.
Capability inventory: Uses Bash(curl) for network access and Bash(python3) for data processing.
Sanitization: The skill uses standard JSON parsing (json.load) to process results, but does not implement additional sanitization for values displayed to the user.

view-usage