The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions specify running a local Node.js script (node scripts/lint.js <file>) to perform quality assurance on generated content. This execution depends on the contents of the scripts/lint.js file, which is not provided in the analyzed context. Use of such scripts is common for automation but represents an execution point for locally stored code.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted data from multiple sources which are then interpolated into generation prompts.
Ingestion points: User input provided as arguments to the /ray-tweet command and local data files such as data/tweets-history.jsonl and queue/*.md.
Boundary markers: The instructions do not define clear delimiters or "ignore embedded instructions" warnings for the ingested data.
Capability inventory: The skill has the capability to execute shell commands via the node interpreter for linting.
Sanitization: There is no evidence of input validation, filtering, or escaping before the external content is processed by the model or passed to the linting script.

ray-tweet