The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions to install the Fly.io CLI using a piped shell script (curl -L https://fly.io/install.sh | sh). Fly.io is a well-known infrastructure service required for the skill's container deployment features. This execution is performed against an official, well-known domain.
[COMMAND_EXECUTION]: The primary purpose of the skill is to execute management commands via npx @insforge/cli. This includes powerful operations such as database schema changes, service lifecycle management (start/stop/delete), and secret manipulation, all of which are intended for administrative use.
[EXTERNAL_DOWNLOADS]: The skill's project creation workflow automatically installs additional agent skills from the vendor's repository (insforge/agent-skills) using the npx skills add command.
[CREDENTIALS_UNSAFE]: The documentation describes the use of environment variables like INSFORGE_PASSWORD for non-interactive authentication. However, it also promotes best practices by providing a dedicated secrets management API (npx @insforge/cli secrets) to avoid hardcoding sensitive values.
[PROMPT_INJECTION]: The diagnose --ai command accepts natural language descriptions of issues as input. This creates a surface for indirect prompt injection:
Ingestion points: User-provided descriptions in the diagnose --ai "<issue>" command (SKILL.md).
Boundary markers: None identified in the instructional prompt.
Capability inventory: The CLI can execute SQL queries, read system logs, manage containerized services, and read decrypted secrets.
Sanitization: No specific sanitization or filtering of the input description is described before it is processed by the diagnostic agent.

insforge-cli