The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill includes an auditing workflow in skills/liquid-glass-auditor.md that directs the agent to scan and read user-provided source files such as .swift, .xcconfig, and .plist. This constitutes a vulnerability surface where instructions maliciously embedded in a project being audited could attempt to influence the agent's behavior.
Ingestion points: The agent is instructed to read source code, property lists, and project configuration files from the local filesystem.
Boundary markers: The auditing instructions do not specify the use of delimiters or 'ignore' warnings when ingesting file content.
Capability inventory: The agent utilizes tool-based file discovery (Glob), searching (Grep), and reading, alongside local command execution for symbol verification.
Sanitization: No explicit sanitization or filtering of the ingested file content is performed before processing.
[COMMAND_EXECUTION]: The skill provides procedures and specific command-line syntax for interacting with the sfsymbols CLI utility, a legitimate Apple tool located within the SF Symbols application bundle. These commands are used to verify symbol availability and rendering support during the design review process.

axiom-design