The Agent Skills Directory

[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill instructs the agent to download an installer script from https://raw.githubusercontent.com/akillness/oh-my-gods/main/install.sh and execute it by piping it directly into bash. This pattern is an extremely high-risk method of software distribution as it executes unverified code from a remote source with the agent's full privileges.
[DYNAMIC_EXECUTION]: The skill facilitates the runtime execution of a shell script downloaded from the internet. It also installs a large catalog of over 80 additional agent skills, which are integrated into the system and potentially executed at runtime, increasing the execution surface area significantly.
[INDIRECT_PROMPT_INJECTION]: By downloading and installing a large volume of external instructions and scripts, the skill creates a significant attack surface where malicious instructions embedded in the external data could influence or override the agent's behavior.
Ingestion points: Remote installer script and the .god-skills/ directory from the akillness/oh-my-gods repository.
Boundary markers: No programmatic boundary markers or automated validation steps are used to isolate the downloaded content.
Capability inventory: The skill is granted powerful capabilities including Bash, WebFetch, Write, and Edit tools, which can be misused if the downloaded content is malicious.
Sanitization: While the instructions suggest an optional "security-conscious" flow for manual human review, the primary and default recommendation is direct execution.
[METADATA_POISONING]: The skill description and compatibility fields in the frontmatter are filled with placeholders (">"), providing no useful information about the skill's actual purpose or environment requirements in the structured metadata.

agenticskills