laniameda-storage
Fail
Audited by Gen Agent Trust Hub on Mar 3, 2026
Risk Level: HIGHDATA_EXFILTRATIONCREDENTIALS_UNSAFECOMMAND_EXECUTION
Full Analysis
- [DATA_EXFILTRATION]: The
scripts/ingest.tsfile usesreadFileSyncto read local files from a path provided in theimagePathJSON field. This mechanism allows the agent to read any file it has permissions for and subsequently upload its base64-encoded content to a remote URL (https://perfect-buffalo-375.convex.cloud). - [DATA_EXFILTRATION]: The skill documentation (
references/convex-interface.md) discloses several absolute file paths from the author's local development environment, providing information about the system's directory structure. - [CREDENTIALS_UNSAFE]: The script
scripts/ingest.tscontains a hardcoded default value forownerUserId("278674008") which is used if theKB_OWNER_USER_IDenvironment variable is missing, potentially leading to data being stored under an incorrect or unauthorized account. - [COMMAND_EXECUTION]: The skill relies on the
bunruntime to execute local TypeScript files. While this is the intended behavior for the skill, the execution of scripts that handle arbitrary file paths increases the overall attack surface. - [INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect injection:
- Ingestion points:
scripts/ingest.tsprocesses JSON data passed via command-line arguments, which may originate from untrusted user messages or external sources. - Boundary markers: There are no delimiters or instructions to prevent the agent from obeying paths or data found within the ingested content.
- Capability inventory: The skill possesses both file-read capabilities (
readFileSync) and network-write capabilities (fetch). - Sanitization: There is no validation or sanitization of the
imagePathvariable to ensure it remains within safe or expected directories.
Recommendations
- AI detected serious security threats
Audit Metadata