The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/run-tests.sh constructs a command string using the $COLLECTION and $ENVIRONMENT variables and executes it via eval. This enables shell command injection if an attacker provides a filename containing shell metacharacters (e.g., ;, &, or backticks).
[REMOTE_CODE_EXECUTION]: Postman collections processed by this skill can contain arbitrary JavaScript code in Pre-request and Test scripts. Newman executes this code at runtime, which means processing an untrusted collection leads to arbitrary code execution in the agent's environment.
[EXTERNAL_DOWNLOADS]: The scripts/install-newman.sh script performs global or local installations of the newman and newman-reporter-htmlextra packages using the NPM registry.
[DATA_EXFILTRATION]: The documentation in references/advanced-patterns.md provides code for a Slack reporter that uses https.request to send test results to external webhooks. This pattern demonstrates how data can be sent to remote endpoints outside the controlled environment.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted Postman collection files without sanitizing or validating the embedded logic.
Ingestion points: Postman collection (.json) and environment files provided by the user or external sources.
Boundary markers: None present to separate collection data from executable scripts.
Capability inventory: The skill possesses the ability to execute shell commands (eval) and perform network requests (via Newman's reporter logic).
Sanitization: No validation of embedded JavaScript content is performed.

newman