The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill accesses sensitive authentication files stored on the local file system.
Evidence: The script upload.mjs defines a credential path at ~/.config/evenhub/credentials.yaml (using homedir() and XDG_CONFIG_HOME).
Behavior: It reads this file to extract access_token and refresh_token for use in API requests.
[DATA_EXFILTRATION]: The skill transmits extracted authentication tokens to external network endpoints.
Evidence: upload.mjs uses fetch to send the access_token in the X-Even-Authorization header to https://hub.evenrealities.com (or a URL defined by the EVENHUB_API_URL environment variable).
Behavior: It also sends the refresh_token in the request body to the /api/v1/auth/refresh endpoint.
[COMMAND_EXECUTION]: The skill modifies sensitive local configuration files.
Evidence: The function writeCreds in upload.mjs uses writeFileSync to overwrite ~/.config/evenhub/credentials.yaml with updated tokens.
[PROMPT_INJECTION]: The skill processes external project files which could be used for indirect injection.
Ingestion points: upload.mjs reads and parses app.json from the current working directory.
Boundary markers: Absent; the script directly interpolates the package_id from app.json into the target API URL.
Capability inventory: The script has network write capabilities (fetch POST) and local file write capabilities (writeFileSync).
Sanitization: None; the package_id is URI-encoded but otherwise used as provided in the local file.

evenhub-upload