The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses npx and node to run the @archastro/astroshot CLI tool for screenshot orchestration and recording. It also supports execution of arbitrary local binaries in PTY mode, as defined in YAML fixtures (e.g., ./target/debug/install-wizard).
[DYNAMIC_EXECUTION]: The React, Ink, and PTY modes dynamically load and execute content from fixtures. React and Ink modes import and render components directly from TypeScript/React files, while PTY mode executes specified local commands with user-defined arguments and working directories.
[REMOTE_CODE_EXECUTION]: The skill fetches and executes the @archastro/astroshot package from the official npm registry using npx. While this targets a well-known service, it involves the execution of code downloaded from a remote repository at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface via fixture files (e.g., fixtures/account-dialog.tsx, fixtures/install-wizard.yaml) which act as ingestion points for untrusted data. There are no automated boundary markers or sanitization processes mentioned beyond the 'Trust boundary' warning in SKILL.md. This capability inventory—including shell execution, local binary execution, and component rendering—could be exploited if an attacker provides malicious fixture content that the agent then processes or executes.

astroshot