The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external and untrusted data, including git commit messages, author names, and user-supplied descriptions. The instructions provide specific mitigation strategies to prevent malicious data from compromising the generated artifact.
Ingestion points: Git repository logs, branch names, and user briefings.
Boundary markers: The skill uses narrative templates to structure content but relies primarily on sanitization rather than delimiters for injection prevention.
Capability inventory: Shell command execution (git), persistent file writing, and tool invocation (frontend-design).
Sanitization: The skill contains explicit instructions in references/guardrails.md and references/recipes/git-promotion-diff.md to entity-encode all embedded text and use textContent instead of innerHTML to prevent Cross-Site Scripting (XSS).
[COMMAND_EXECUTION]: The skill uses local shell commands, primarily git and common utilities like awk, sort, and grep, to gather data for the 'git branch promotion diff' recipe. These commands are executed locally to analyze the repository state.
[EXTERNAL_DOWNLOADS]: The skill references resources from well-known and trusted services to provide styling and typography for the generated HTML artifacts.
Resources: Tailwind CSS v4 via JSDelivr (cdn.jsdelivr.net), Google Fonts (fonts.googleapis.com), and Mermaid diagrams via CDN.
Tone: These references are standard for web-based artifact generation and target official, well-known providers.

artifact