The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from GitHub Pull Requests (titles, labels, and descriptions) which are controlled by external users. This data is processed and aggregated into reports, creating a surface for indirect prompt injection if the output is consumed by other LLM-based agents.
Ingestion points: PR metadata fetched via the gh CLI in scripts/generate-report.js.
Boundary markers: The scripts/generate-report.js script includes an escapeHtml function to sanitize PR titles before including them in HTML templates.
Capability inventory: The skill can write files to the local system (fs.writeFileSync) and execute system commands (execFileSync, shell scripts).
Sanitization: Active HTML escaping is used to prevent XSS and direct injection in the generated report artifacts.
[COMMAND_EXECUTION]: The skill uses local command execution to interact with the environment and generate deliverables.
scripts/generate-report.js executes the GitHub CLI (gh) to retrieve data.
scripts/html-to-pdf.sh invokes local browsers (google-chrome, chromium) and conversion tools (wkhtmltopdf) to export reports.
scripts/puppeteer-pdf.js uses puppeteer to launch a headless browser for rendering.
[EXTERNAL_DOWNLOADS]: The report generation template (templates/client-report.html) fetches the chart.js library from cdn.jsdelivr.net. This is a well-known and trusted Content Delivery Network service.

launch