The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends running npx gitnexus analyze to update the search index. This command downloads and executes the gitnexus utility from the npm registry.
[COMMAND_EXECUTION]: Users are instructed to run npx gitnexus analyze in their terminal to resolve stale index errors. This provides a mechanism for local command execution to perform repository analysis.
[PROMPT_INJECTION]: The skill processes potentially untrusted external data (error logs and source code) to identify bugs, creating an attack surface for indirect prompt injection.
Ingestion points: Error messages and symptom descriptions passed to gitnexus_query in SKILL.md, as well as source code content read during the trace.
Boundary markers: No specific delimiters or safety instructions are used in the skill to separate untrusted data from agent instructions.
Capability inventory: The agent instructions include suggesting shell command execution via the terminal tool.
Sanitization: There is no evidence of sanitization or validation of the input data before it is processed by the agent.

gitnexus-debugging