The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill features a specialized renderer for GitHub issues (scripts/render-github-issues.py) that fetches external data and embeds it into a generated HTML report. This ingestion point creates an attack surface where malicious instructions in issue titles or comments (untrusted data) could attempt to influence the agent's behavior during subsequent processing of the report. The implementation mitigates traditional injection by using DOM APIs (textContent) for rendering and JSON-encoding with escaped closing tags to ensure data stays contained within its script container.
[DYNAMIC_EXECUTION]: Several orchestration and validation scripts use importlib.import_module() to load local utility files (e.g., loading css_slop_rules in scripts/validate-artifact.py). While dynamic, these imports utilize hardcoded strings and restricted local paths, minimizing risk.
[COMMAND_EXECUTION]: The skill frequently uses subprocess.run() to interact with the GitHub CLI, LibreOffice, and Playwright for artifact generation and export. These operations are implemented securely using argument lists rather than shell strings, preventing command injection. Additionally, generated filenames are restricted to a safe alphanumeric kebab-case format.

html-artifact