The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is designed to read highly sensitive financial information, including balances.json, transactions.json, investments.json, and mortgage.json. While the skill's primary purpose is to generate a dashboard, it inlines the source data into a static HTML file at output_path. This creates a risk where users may unknowingly share their entire financial history if the generated HTML file is transmitted or hosted insecurely.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data from sources like transaction descriptions and alert feeds to populate the dashboard. There is no explicit requirement for HTML sanitization or output escaping in the instructions. This creates a vulnerability surface where malicious strings in the source JSON (e.g., a transaction named <script>alert(1)</script>) could result in Stored Cross-Site Scripting (XSS) within the generated HTML report.
Ingestion points: The skill reads transactions.json, goals.json, and recent files in reports/alerts/ (specified in SKILL.md).
Boundary markers: No boundary markers or "ignore instructions" delimiters are specified for the data interpolation.
Capability inventory: The skill has the capability to write to the local file system (output_path) and assemble HTML/JavaScript content.
Sanitization: While privacy masking for account numbers is mentioned, no general escaping or validation of user-controlled text fields is required during the rendering phase.
[EXTERNAL_DOWNLOADS]: The skill references an external JavaScript library (D3.js) via the JSDelivr CDN (https://cdn.jsdelivr.net/npm/d3@7). The skill correctly specifies the use of integrity hashes (SRI) to ensure the integrity of the remote resource, which is a security best practice.

household-finance-dashboard-builder