The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external web pages, creating a surface for indirect prompt injection where malicious page content could influence agent behavior.
Ingestion points: JavaScript profiling scripts in references/browser-profiling.md extract DOM elements, text content, and class names from the active browser tab.
Capability inventory: The skill is authorized to modify source code, execute build scripts via npm, and perform git commits.
Boundary markers: No explicit delimiters or instructions are provided to isolate untrusted web content from the agent's control logic.
Sanitization: Captured text data is not sanitized for malicious instructions before being processed by the agent.
[COMMAND_EXECUTION]: The skill uses shell commands to interact with the local repository and validate builds.
Evidence: The SKILL.md file instructs the agent to run git status, git diff, npm run build, and npm run lint. It also describes a method for creating isolated git commits using GIT_INDEX_FILE and temporary index files.
[DYNAMIC_EXECUTION]: The skill executes dynamic JavaScript within the browser environment to perform technical audits.
Evidence: references/browser-profiling.md provides templates for functions like profileAnimationPage and samplePerformancePage which are injected into the browser context using playwright.evaluate to measure performance metrics.

optimize-web-animations