The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill suggests installation via npx from the author's repository and utilizes npx to fetch and run chrome-devtools-mcp@latest and lighthouse for data collection.
[COMMAND_EXECUTION]: The skill executes shell commands such as npx for performance auditing and grep for searching the local workspace to map bottlenecks to source code. It also includes functionality to implement code changes when requested.
[PROMPT_INJECTION]: The skill is subject to an indirect prompt injection surface as it processes external content like URLs, performance traces, and screenshots.
Ingestion points: Target URLs, performance trace files (JSON), and flame graph images (processed via vision).
Boundary markers: No explicit delimiters are specified to isolate untrusted trace data from the agent's instructions.
Capability inventory: Access to shell execution (npx), file system search (grep), and file modification capabilities.
Sanitization: The instructions do not define specific validation or sanitization steps for the ingested performance artifacts.

flame-audit