bankr-communities
Fail
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: CRITICALEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONPROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references the domain
https://qrcoin.funas a platform for QR auctions inSKILL-LINKED-FUNDRAISERS.md. This domain has been flagged as blacklisted by automated security scanners, posing a risk to users interacting with linked fundraisers. - [CREDENTIALS_UNSAFE]: The skill facilitates the handling of
bk_...API keys for synchronizing projects between Bankr Space and Bankr Agent Profiles. While the instructions inreferences/BANKR-API-KEYS.mdemphasize secure handling (DM-only submission, no logging, and server-side verification), the agent is explicitly tasked with processing these sensitive credentials. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, including X (formerly Twitter) status updates, space posts, and external API response strings (
replyText,instruction). Although the skill provides comprehensive defensive formatting rules inreferences/RESPONSE-SAFETY.md, the reliance on user-controlled content from social media platforms creates a vulnerability surface for indirect prompt injection. - [COMMAND_EXECUTION]: Documentation for the "Platform Agent" and "0xWork" integrations includes instructions for executing shell commands, such as using
curlfor API interactions and runningnpm install -g @0xwork/clito install command-line tools. - [PROMPT_INJECTION]: The file
references/PROMPT-INJECTION.mdcontains strings and patterns typical of prompt injection attempts (e.g., "ignore previous rules"). While these are documented as examples of untrusted content for the agent to disregard, their presence triggered automated detection for high-severity prompt injection patterns. - [METADATA_POISONING]: Automated reputation scans flagged
SKILL.mdwith a malicious file reputation (FileRepMalware). This is likely due to the inclusion of blacklisted URLs and sensitive cryptocurrency-related terminology within the skill's instructions and metadata.
Recommendations
- CRITICAL: 1 file(s) identified as malware by FileRep - DO NOT USE
- Contains 1 malicious URL(s) - DO NOT USE
Audit Metadata