The Agent Skills Directory

[DATA_EXFILTRATION]: The skill extracts information from "cached timeline data" and facilitates its upload to GitHub Discussions or Gists on the ww-w-ai/super-token-saver repository. While this functionality supports the skill's goal of mapping rate limit formulas and targets the vendor's own infrastructure, users should be cautious as timeline caches may contain sensitive prompt or response data.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface where it ingests untrusted data from an external website to modify its internal files.
Ingestion points: The skill uses WebFetch to retrieve model pricing information from platform.claude.com (Anthropic's documentation).
Boundary markers: There are no explicit delimiters or "ignore embedded instructions" warnings applied to the fetched data before processing.
Capability inventory: The skill can execute local Node.js scripts and modify files within its own directory structure (e.g., scripts/model-pricing.json).
Sanitization: The instruction requires confirmation of six specific data fields before editing, which provides basic structural validation but does not prevent logic-based injection if the remote source were compromised.
[DYNAMIC_EXECUTION]: The skill performs self-modification of its own project files. When it encounters an unknown model, it is instructed to edit scripts/model-pricing.json at runtime using data parsed from a remote WebFetch. This dynamic updating of the skill's own configuration assets from external data sources presents a risk of unintended behavior or configuration corruption.
[COMMAND_EXECUTION]: The skill executes a local Node.js script (report-limit.js) via the shell, passing user-supplied arguments such as plan types and dates to the command line.

report-limit