etsy-tony-full-erank-tools
Pass
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: SAFECREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill provides commands such as
config set etsy_api_keyto store Etsy API keys and OAuth tokens in a local session file named.erank_session.json(managed inharness/cli_anything/erank/core/session.py). Although the CLI tool masks these values during display, the sensitive credentials are saved in plaintext within the local JSON file on the user's filesystem. - [COMMAND_EXECUTION]: The wrapper script
scripts/run_erank.pyutilizessubprocess.runto execute the local Python modulecli_anything.erankusing user-supplied arguments. This pattern is also present in the End-to-End (E2E) testing suite to validate CLI functionality. - [EXTERNAL_DOWNLOADS]: The harness performs automated network requests to fetch marketplace data from official and third-party sources, including
openapi.etsy.com,alura.io, and undocumented endpoints atmembers.erank.com/api(e.g.,/api/top-sellersand/api/shop-info). These requests use standard HTTP methods and specific headers to retrieve JSON data for research purposes. - [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection due to its processing of external marketplace data.
- Ingestion points: The skill reads untrusted data from local CSV and JSON exports (e.g., listing titles, descriptions, and tags) via
load_listingsinharness/cli_anything/erank/utils/data_io.py. - Boundary markers: Minimal boundary markers or 'ignore' instructions are present when this data is processed for audits.
- Capability inventory: The skill has the capability to write files (
write_csv,write_json) and execute subprocesses via the harness wrapper. - Sanitization: There is no evidence of specific sanitization or escaping of external content before it is interpolated into analysis reports or audits.
- [METADATA_POISONING]: The
SKILL.mdfile contains promotional metadata and instructions in Chinese directing users to a Douyin account (ID: 93440780745) for support or demonstrations. While not malicious, this represents promotional metadata embedded within the skill instructions.
Audit Metadata