The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from source files and rendered web pages to perform UI audits.
Ingestion points: The scripts/scan-slop.mjs script reads local source files, and scripts/_shared/layout-probe.mjs visits external URLs using Playwright.
Boundary markers: The skill provides detailed operating rules in references/_shared/experience-core.md and a references/final-slop-gate.md to ensure the agent follows evidence-based reasoning rather than arbitrary content.
Capability inventory: The tools extract technical signals (e.g., CSS properties, text patterns) used for quality verdicts.
Sanitization: The scanning logic uses fixed regular expressions to count specific signals rather than interpreting the content as instructions.
[DYNAMIC_EXECUTION]: The browser-based inventory tool in scripts/_shared/layout-probe.mjs uses page.evaluate() to execute predefined data collection scripts within a browser context.
[EXTERNAL_DOWNLOADS]: The skill's layout scanner relies on Playwright, a well-known browser automation framework from Microsoft. The installation instructions in scripts/_shared/layout-probe.mjs refer to official package registries and browser binaries.

anti-slop-ui