The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill implements secure credential management practices. It instructs users to store sensitive OAuth 2.0 credentials in a local .env file and provides specific commands to add this file to .gitignore to prevent accidental credential exposure in version control systems.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of standard, widely-used libraries from official registries, including dotenv for Node.js and requests and python-dotenv for Python. These are established tools for handling environment variables and HTTP operations.
[REMOTE_CODE_EXECUTION]: An example command demonstrates using curl to fetch a token from Navan's API and piping the result to python3 for JSON parsing. As the target is the official API of a well-known service (api.navan.com), this is considered a safe integration pattern for verification purposes.
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it ingests data from the Navan API and processes user-supplied credentials. However, the logic applied to this data (confirming token acquisition and checking record counts) does not present an exploitable capability chain.
Ingestion points: Local .env file configuration and remote responses from the Navan REST API.
Boundary markers: None explicitly defined in the code snippets, relying on standard JSON deserialization.
Capability inventory: Includes network requests (fetch, requests, curl), file system writes (for .env and .gitignore), and package management (npm, pip).
Sanitization: Uses standard platform libraries for environment variable parsing and HTTP response handling.

navan-install-auth