The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill analyzes untrusted trace files which could contain malicious instructions embedded in metadata fields like span names.
Ingestion points: scripts/analyze_trace.js (reads user-provided trace.json files).
Boundary markers: Absent; the analysis output lacks delimiters or warnings to ignore embedded instructions.
Capability inventory: The agent is instructed to read local configuration files and source code based on the analysis results, creating an exploitation path.
Sanitization: Absent; the analysis script directly prints the contents of JSON fields like span.name and fields.resource.
Command Execution (SAFE): The workflow involves running standard build and diagnostic commands such as pnpm build, node, and tail. These are appropriate for the primary purpose of the skill and do not involve unsanitized user input in command arguments.

rspack-tracing