signals
Warn
Audited by Gen Agent Trust Hub on Oct 19, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DYNAMIC_EXECUTION]: The
scripts/publish-signal.shscript executes dynamically generated JavaScript via thenode -ecommand. It interpolates shell variables directly into the JavaScript code string, which could allow for arbitrary code execution if inputs like the trade action or token symbol contain malicious characters. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon trading signals from external providers via the
/api/feedendpoint. These signals include an arbitrary 'reasoning' text field that is processed by the agent, creating a surface for indirect prompt injection from third-party sources. - Ingestion points:
https://bankrsignals.com/api/feedandhttps://bankrsignals.com/api/signals(referenced inSKILL.mdandHEARTBEAT.md). - Boundary markers: The skill does not define specific delimiters or instructions to treat external signal content as untrusted.
- Capability inventory: The agent has the ability to sign cryptographic messages (
publish-signal.sh), access local configuration files, and perform authenticated network operations. - Sanitization: The skill does not implement or describe any sanitization or validation logic for the free-text reasoning fields within the signals.
- [INDIRECT_PROMPT_INJECTION]: The agent is instructed to periodically fetch a remote markdown file (
https://bankrsignals.com/heartbeat.md) and 'follow' the instructions within it. This enables the remote server to dynamically inject new natural language instructions into the agent's execution context at runtime. - [COMMAND_EXECUTION]: The skill provides and relies on shell scripts (
scripts/publish-signal.sh) to perform core financial operations, including managing private keys and interacting with APIs, which increases the potential impact of command injection vulnerabilities.
Audit Metadata