predict-intelligence
PassAudited by ClawScan on May 1, 2026.
Overview
This skill appears coherent for generating prediction-report PDFs, with disclosed web access, local file/Python execution, and third-party dependencies that users should review.
Before installing, be comfortable with the agent browsing the web, writing local report files, running Python/Playwright to create PDFs, and loading visualization resources from third-party CDNs. Use a virtual environment and review generated reports if they contain private or decision-sensitive information.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
When invoked, the agent may browse the web, create report files, and run the bundled Python conversion scripts.
The skill explicitly requests browsing, URL fetching, local file writes, and shell execution. These capabilities are expected for generating PDF reports, but they give the agent meaningful local and network authority.
| **Web search** | Search the internet for news, analysis, data | | **URL fetch** | Open a URL and read its content | | **File read/write** | Read template, write HTML report | | **Shell execute** | Run Python 3.9+ scripts |
Use it for intended report-generation tasks, and review any shell command or output path before allowing execution.
If you install the optional requirements file, your environment will pull multiple third-party Python packages in addition to Playwright.
The included requirements file uses external packages with version lower bounds rather than exact pins. This is common, and the main documented setup is user-directed, but installing optional dependencies expands the supply-chain surface.
jinja2>=3.1 requests>=2.31 staticmap>=0.5.7 Pillow>=10.0 playwright>=1.40
Install only the dependencies needed for the workflow you plan to use, preferably in a virtual environment, and review dependency versions if reproducibility matters.
Rendering a report may execute scripts embedded in the generated HTML and load resources needed for visualizations.
The converter opens the generated HTML file in Chromium to render it as a PDF. This is necessary for D3-based reports, but it means JavaScript in the report page can run during rendering.
page.goto(f"file://{abs_html}", wait_until="networkidle", timeout=timeout * 1000)
page.wait_for_timeout(5000)
page.pdf(Render only reports generated for the current task, avoid inserting raw untrusted HTML into reports, and be cautious with confidential content when remote resources are loaded.