The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads the official CLI installer and setup scripts from the vendor's repository at raw.githubusercontent.com/InsForge/insta-cli/main/install.sh and agents.instacloud.com. These are legitimate resources owned by the skill author used for environment initialization.
[REMOTE_CODE_EXECUTION]: The skill executes the downloaded setup script by piping the output of curl directly to the shell (sh). This is the intended installation and upgrade mechanism for the insta toolset.
[COMMAND_EXECUTION]: The skill makes extensive use of shell commands to drive the insta CLI, manage project files, and perform data operations. It also utilizes standard utilities like git for repository management and psql/pg_dump for database migration and verification.
[INDIRECT_PROMPT_INJECTION]: The skill contains logic for processing untrusted external data, primarily during the migration of applications from other platforms or when deploying service templates from remote GitHub URLs.
Ingestion points: External data enters the context when the agent connects to third-party database URLs for migrations or clones remote repositories for template deployments.
Boundary markers: The instructions rely on a robust "approval relay" system where sensitive mutations require an admin to provide a single-use authorization code, ensuring human oversight of the agent's actions.
Capability inventory: The agent can execute remote commands on compute instances, manage environment secrets, and perform network requests to verify service health.
Sanitization: The skill emphasizes the use of environment variables and shell pipes to handle sensitive credentials securely, preventing their exposure in command logs or stdout. It also includes the insta --agent observe tool to detect and redacted sensitive information in tool outputs.

insta