The Agent Skills Directory

[SAFE]: The skill performs legitimate design verification and auditing tasks using standard platform tools. There are no signs of malicious logic, obfuscation, or unauthorized data access.
[EXTERNAL_DOWNLOADS]: The skill fetches design context, screenshots, and metadata from Figma's official platform via authorized MCP tools. As Figma is a well-known design service and these operations are fundamental to the skill's purpose, they are considered safe.
[COMMAND_EXECUTION]: Provides predefined shell commands using grep to audit the local codebase for temporary design placeholders or pending designer confirmations. These commands are transparent, read-only, and do not present a security risk.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external design specifications and browser snapshots, creating a surface for potential indirect prompt injection where instructions could be hidden in design files.
Ingestion points: Step 2 and Step 3 in SKILL.md (via Figma MCP and Playwright browser snapshots).
Boundary markers: No specific delimiters or markers are specified to distinguish external design content from instructions.
Capability inventory: The agent has access to Bash, Read, and Playwright tools, allowing for file system and browser access.
Sanitization: No explicit sanitization or filtering of design text or variable values is implemented.

figma-verify