The Agent Skills Directory

[CREDENTIALS_UNSAFE] (HIGH): Hardcoded Terra API credentials for multiple environments including production.
Evidence: The file SKILL.md contains explicit dev_id and api_key values for testing, staging, and production environments (e.g., botaniqalmedtech-prod-yrc99AwsT2).
Impact: Anyone with access to the skill can authenticate as the 'botaniqalmedtech' entity, potentially accessing sensitive health data or modifying integrations.
[DATA_EXFILTRATION] (HIGH): The skill documentation encourages and provides code for accessing sensitive local credential files.
Evidence: Code snippets explicitly read from .env.terra.production and .env.terra.staging using load_dotenv.
[INDIRECT_PROMPT_INJECTION] (MEDIUM): The skill exposes an attack surface by accepting unvalidated external URLs and reference IDs that are passed to the Terra API.
Ingestion points: generate_widget_session function in SKILL.md.
Boundary markers: None present.
Capability inventory: Makes network requests to api.tryterra.co via the terra SDK.
Sanitization: No evidence of URL validation or input escaping before passing data to the SDK.

terra-auth