The Agent Skills Directory

[SAFE]: The skill defines a secure workflow for generating static HTML artifacts. It explicitly mandates that all files must be self-contained, using only inline CSS and local-only JavaScript. It contains a comprehensive 'Safety checklist' that strictly forbids external scripts, remote fonts, network requests, analytics, and the inclusion of credentials or secrets.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data (such as code diffs and research reports) to generate its artifacts. While this creates a potential injection surface, the skill implements strong procedural guardrails.
Ingestion points: Workflow Step 2 in SKILL.md requires reading project files, diffs, and research.
Boundary markers: The skill relies on structured templates (template.html) rather than raw interpolation, reducing the risk of accidental execution.
Capability inventory: The skill leverages the agent's ability to write files to the artifacts/html/ directory.
Sanitization: The 'Safety checklist' explicitly directs the agent to scan the final output for <script src> tags, tracking pixels, and fetch calls, providing a critical verification step before the artifact is provided to the user.
[DYNAMIC_EXECUTION]: The template.html file includes a minimal, inline JavaScript function for local clipboard interaction. This script is strictly scoped to local functionality and does not facilitate remote code execution or network communication.

html