The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill attempts to execute the solograph package using the command uvx solograph --help. The uvx tool downloads and executes code from the Python Package Index (PyPI) at runtime. As solograph is not associated with a trusted organization or the skill author ('fortunto2'), this represents the execution of untrusted remote code.
[EXTERNAL_DOWNLOADS]: The skill triggers the download of the solograph package from an external registry (PyPI) during the tool verification step.
[COMMAND_EXECUTION]: The skill uses the Bash tool for system-level operations, including directory creation (mkdir -p ~/.solo-factory) and the execution of the uvx check.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection by incorporating unsanitized user responses and external template content into markdown and YAML configuration files. These files (e.g., manifest.md, dev-principles.md) are designed to be consumed by other skills, creating an attack surface where malicious user input could influence subsequent agent behavior.
Ingestion points: User answers gathered via AskUserQuestion and local template files from the templates/ directory.
Boundary markers: No delimiters or warnings are present in the generation logic to isolate user-provided content within the generated configuration files.
Capability inventory: The skill utilizes Bash, Write, and Edit to configure and persist the project environment.
Sanitization: The skill does not implement input validation or escaping for user-provided strings before writing them to persistent files.

solo-init