debug

Pass

Audited by Gen Agent Trust Hub on Nov 19, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill instructions require the agent to capture and process raw error output, stack traces, and build logs verbatim. This creates a significant surface for indirect prompt injection, where malicious content embedded in logs or code could potentially influence the agent's behavior.
  • Ingestion points: Raw error output from build systems, webviews, and runtime stack traces (SKILL.md).
  • Boundary markers: Absent. The instructions explicitly call for capturing "raw error output verbatim" without recommending the use of delimiters or instructions to ignore embedded commands.
  • Capability inventory: The skill possesses write access to configuration files (.vscode/launch.json), the ability to invoke VS Code commands, and the ability to execute local shell scripts (scripts/error-detection.sh/ps1) (SKILL.md).
  • Sanitization: Absent. The workflow prioritizes preserving "full raw error telemetry" and original error output for analysis.
  • [DYNAMIC_EXECUTION]: The skill programmatically generates and updates the .vscode/launch.json file. Because this configuration file defines how programs are launched, including environment variables and program arguments, its dynamic generation poses a risk if the input data used to construct the configuration is influenced by malicious code or untrusted error logs.
  • [COMMAND_EXECUTION]: The agent is instructed to execute local scripts (scripts/error-detection.ps1 on Windows or scripts/error-detection.sh on macOS/Linux) to parse build errors and match them against known issues. The execution of these scripts occurs as part of the automated diagnostic loop.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 19, 2026, 01:17 AM