openclaw-reportstudio-community

Generate polished, read-only business reports from CSV/XLSX into static deliverables (xlsx+pdf+pptx) using ReportStudio Community. Use when a user says things like “用这个表生成月报/周报/经营分析/PPT+PDF汇报材料”, provides a file path, asks for trends (day/week/month), TopN breakdowns by a dimension/metric, or wants professional exports without hallucinated claims.

This skill appears to do exactly what it says: run a local ReportStudio CLI on a CSV/XLSX and return artifacts. The primary remaining risk is that it will execute whatever 'reportstudio' code is available in the Python environment — that code could perform network requests, access other files, or be malicious if you installed an untrusted package. Before installing/using: (1) ensure the reportstudio package you rely on is from a trusted source or run the skill in an isolated virtualenv/container, (2) avoid passing sensitive files unless you have verified ReportStudio's behavior, (3) do not point repo_dir at untrusted locations, and (4) run the acceptance checklist (references/acceptance.md) and an interactive python -m reportstudio.cli.main --help to confirm behavior. If you need stronger guarantees, require network egress controls or run in a locked-down sandbox.

Type: OpenClaw Skill Name: openclaw-reportstudio-community Version: 1.0.0 The `SKILL.md` file instructs the AI agent to execute a `python -m` command, providing an example where user-controlled parameters like `<PATH>`, `<OUT_DIR>`, `<DIM_COL>`, and `<MEASURE_COL>` are passed without quotes. This creates a shell injection vulnerability if the agent directly constructs and executes this command in a shell. Although a safer wrapper script (`scripts/run_reportstudio.py`) exists that uses `subprocess.run` with a list (mitigating shell injection), the `SKILL.md` states its use is 'optional' and to 'prioritize running directly', implying the agent might not use the wrapper. This represents a critical vulnerability (RCE risk) but lacks clear evidence of malicious intent, as the skill explicitly claims 'read-only' and 'no default external network access'.