The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute a shell command node .gitnexus/run.cjs analyze in the terminal to refresh the index. This involves executing a script file located within the repository's directory.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to search, read, and analyze source code and error symptoms, which are untrusted inputs.
Ingestion points: Repository source code, error messages, and execution traces accessed via query, context, and gitnexus:// resources.
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potentially malicious instructions embedded within the analyzed code.
Capability inventory: The agent can execute shell commands (node), perform Cypher queries, and read arbitrary resources from the repository.
Sanitization: There is no evidence of sanitization or filtering of the content retrieved from the repository before it is processed by the agent.
[DYNAMIC_EXECUTION]: The instruction to run node .gitnexus/run.cjs analyze triggers the execution of a Javascript file (run.cjs) found in the project's hidden .gitnexus directory. If the agent is used to analyze an untrusted repository, this script could be modified by an attacker to execute arbitrary code.

gitnexus-debugging