optimize-web-animations
Pass
Audited by Gen Agent Trust Hub on Nov 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The profiling logic in
references/browser-profiling.mdextracts text content from DOM elements usingelement.textContentto help identify owners of animations. - Ingestion points: Data extracted from web pages via browser evaluation is returned directly to the agent's context in
references/browser-profiling.md. - Boundary markers: No explicit delimiters or instructions to ignore embedded content are used when presenting the extracted text to the agent.
- Capability inventory: The skill has access to sensitive tools including
Bash,Write, andEdit, which could be exploited if the agent obeys instructions found within the processed web data. - Sanitization: The skill does not perform sanitization or filtering on the text content captured from the browser.
- [DYNAMIC_EXECUTION]: The skill relies on executing dynamic JavaScript snippets within the browser context and a
node_replenvironment to perform page audits. - Implementation: It uses
tab.playwright.evaluateto run profiling scripts defined inreferences/browser-profiling.mdand usesnode_replto import and set up the browser client. - [COMMAND_EXECUTION]: The workflow involves executing several shell commands to inspect the environment and verify builds.
- Evidence: The skill uses
git status --short,npm run lint, andnpm run buildas part of its standard verification and inspection process.
Audit Metadata