gangtise-kb
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The HTTP client in
scripts/_client.pyexplicitly disables SSL certificate validation and hostname checking, creating a risk during credential transmission.\n - Evidence: The script sets
_ssl_ctx.check_hostname = Falseand_ssl_ctx.verify_mode = ssl.CERT_NONEwhen initializing the SSL context used for all API communication withopen.gangtise.com.\n - Risk: This implementation allows for Man-in-the-Middle (MITM) attacks, which could be used to intercept sensitive API keys (
GANGTISE_ACCESS_KEYandGANGTISE_SECRET_KEY) or manipulate the financial data returned by the service.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external research reports and meeting summaries, which could be exploited for indirect prompt injection.\n - Ingestion points: The skill fetches data from
open.gangtise.comusing scripts likescripts/query_kb.py,scripts/meeting_list.py, andscripts/indicator.py.\n - Boundary markers: There are no delimiters or specific instructions to the agent to ignore embedded commands within the tool outputs returned in
mcp-server/server.py.\n - Capability inventory: The skill has capabilities for network communication and file system writes (via
scripts/download_resource.py).\n - Sanitization: While
scripts/clean_md.pyremoves HTML tags, it does not sanitize the text content for malicious instruction-like patterns.\n- [EXTERNAL_DOWNLOADS]: The skill provides functionality to download external resources to the local file system.\n - Evidence:
scripts/download_resource.pycontains logic to download PDF reports and binary data from the Gangtise API and save them locally.\n - Risk: The script determines local filenames using the
Content-Dispositionheader from the server without robust path validation, which could theoretically be exploited if the source domain were compromised.\n- [COMMAND_EXECUTION]: The skill instructions encourage the execution of local setup scripts for configuration.\n - Evidence:
SKILL.mdandmcp-server/README.mddirect users to runpython3 scripts/configure.pyto manage API credentials.
Audit Metadata