The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell pipelines using ccusage, jq, and termgraph. It relies on the Bash tool to orchestrate data processing and visualization based on user-supplied arguments.
[EXTERNAL_DOWNLOADS]: Fetches and runs the termgraph package at runtime via the uvx package runner from well-known sources.
[DATA_EXFILTRATION]: Accesses a specific configuration and log directory (/Users/alexei/.claude-team-gaia-mbp-m2) to extract token and cost data. This involves reading from a potentially sensitive absolute filesystem path.
[PROMPT_INJECTION]: Subject to indirect prompt injection (Category 8) because it processes untrusted log data that may contain malicious instructions.
Ingestion points: JSONL logs are ingested via the ccusage tool (file: SKILL.md).
Boundary markers: None present in the data processing flow.
Capability inventory: Shell command execution via Bash and remote package execution via uvx.
Sanitization: Employs jq for data extraction and rounding, which does not validate against or strip natural language instructions within the logs.

analyzing-usage