The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/generate_report.py uses subprocess.run to invoke a local rendering tool (rq-report-renderer) to convert Markdown reports into HTML. The execution uses a fixed binary name and argument list without a shell environment, minimizing the risk of command injection.
[EXTERNAL_DOWNLOADS]: In scripts/extract_announcements.py, the skill fetches official announcement PDF documents from well-known financial domains, including the Shanghai Stock Exchange (sse.com.cn), Shenzhen Stock Exchange (szse.cn), and CNINFO (cninfo.com.cn). These operations are consistent with the skill's intended purpose of financial analysis.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it processes untrusted data from web search results and external research reports and interpolates it into the report context.
Ingestion points: Untrusted data enters the context through web_search_findings.json and research_reports.json.
Boundary markers: Absent. The data is directly interpolated into Markdown templates without specialized delimiters or instructions to ignore embedded commands.
Capability inventory: The skill has the capacity to execute system commands via the report renderer and perform file system writes.
Sanitization: The processing logic includes whitespace normalization and string truncation but does not escape Markdown control characters or instructions that might be embedded in the external content.

rq-earnings-preview