The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the node runtime to execute a local script located at scripts/trace_query.bundle.cjs. This file is referenced throughout the SKILL.md instructions but is not included in the provided 9 files, making it an unverifiable executable dependency.
[EXTERNAL_DOWNLOADS]: The trace query tools (such as id, metadata, and sql) support a --path option that can target remote URLs (e.g., https://example.com/trace.pftrace), allowing the skill to fetch and process data from arbitrary external servers.
[COMMAND_EXECUTION]: The sql tool allows the execution of raw SQL queries against trace data. While standard for performance analysis, this provides a powerful mechanism for data extraction and manipulation within the agent's environment.
[PROMPT_INJECTION]: The skill ingests untrusted trace data from external URLs and processes it using tools with significant capabilities, which constitutes an indirect prompt injection surface.
Ingestion points: Trace files and metadata loaded from external URLs or local paths via the --path argument (SKILL.md).
Boundary markers: The skill does not implement boundary markers or instructions to disregard embedded commands in the trace data.
Capability inventory: Execution of shell commands via node, raw SQL query execution, and the ability to write files to disk using the readData tool with the -o output flag (SKILL.md).
Sanitization: There is no evidence of content validation or sanitization for the data contained within the trace files prior to analysis.

lynx-trace-analysis