lightrun-live-runtime-debugging
Pass
Audited by Gen Agent Trust Hub on Jul 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill connects to the official Lightrun MCP server at
app.lightrun.comand references troubleshooting documentation ondocs.lightrun.com. These are recognized as legitimate vendor resources.\n- [DATA_EXFILTRATION]: Usage tracking is implemented via a call to askill_starttool, providing only the skill name for telemetry. This is a standard practice for usage monitoring and does not involve sensitive user data or environment secrets.\n- [COMMAND_EXECUTION]: The skill instructs the agent to manage debugging actions, such as creating, polling, and canceling probes, through the Lightrun MCP interface. These operations are restricted to the authorized debugging platform and follow a structured workflow.\n- [PROMPT_INJECTION]: Surface analysis for indirect prompt injection reveals that the skill ingests runtime data (captured values and call stacks) from the Lightrun MCP and has capabilities to further interact with the debugging tools. This represents a known risk surface inherent to debugging tasks.\n - Ingestion points: Captured runtime evidence (values, stack traces) from Lightrun MCP tools (documented in SKILL.md and references/mcp-tool-discovery.md).\n
- Boundary markers: None explicitly mentioned in the skill instructions for segregating runtime data from instructions.\n
- Capability inventory: Lightrun MCP tools for managing runtime actions (creating, checking status, retrieving data, canceling) as described in references/mcp-tool-discovery.md.\n
- Sanitization: No specific escaping or validation logic is defined for the captured runtime content within the skill's instructions.
Audit Metadata