The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill instructs the agent to create new test files (e.g., screenshots-adhoc.spec.ts) and execute them using the local runner, which involves generating and running code at runtime.\n- [REMOTE_CODE_EXECUTION]: The workflow involves running local project tests and scripts via pnpm and node. While no remote scripts are downloaded, the agent executes existing code within the project workspace that could be compromised if the repository content is untrustworthy.\n- [DATA_EXPOSURE]: The skill reads local screenshot files, converts them to Base64, and uploads them to a remote GitHub branch using the Git Data API. This establishes a mechanism for exporting local data to external infrastructure.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted project data that could influence its behavior.\n
Ingestion points: UI source files and test specifications (e.g., screenshots.spec.ts) referenced in SKILL.md.\n
Boundary markers: None identified; instructions do not include specific delimiters or warnings to ignore embedded commands in the source files.\n
Capability inventory: Subprocess calls for pnpm, cargo, and gh api are used to manage the workspace and network state.\n
Sanitization: No explicit sanitization or filtering of UI content or fixture data before processing.

lite-screenshots