The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted repository content to determine project dependencies and setup requirements, which presents a surface for indirect prompt injection.
Ingestion points: The agent inspects local repository files, lockfiles, and project structure in the 'Inspect the repo' phase.
Boundary markers: Absent; there are no instructions to use delimiters or ignore embedded instructions within the files being analyzed.
Capability inventory: The agent can author shell scripts (.superset/setup.sh) and execute CLI tools (superset workspaces create) based on its findings.
Sanitization: Absent; the skill does not specify validation or filtering of the repository data before it influences script generation.
[COMMAND_EXECUTION]: The skill uses the 'superset' CLI to create and verify workspaces, which involves running shell commands defined in the authored configuration files.
[DYNAMIC_EXECUTION]: The core functionality involves generating executable shell scripts (setup.sh, teardown.sh) and a lifecycle configuration (config.json) that are executed by the platform's workspace manager. This is the primary intended purpose of the skill and includes a mandatory step to 'Show the user the proposed files and get explicit approval before writing.'
[DATA_EXPOSURE]: The skill instructions involve handling sensitive environment files (.env). It explicitly follows best practices by instructing the agent to copy these files locally within the environment and specifically warns to 'never commit them', reducing the risk of accidental credential exposure.

superset-setup