The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands via the Bash tool to interact with the OpenAnt platform. User-provided strings for task titles and descriptions are interpolated into these commands, which could be exploited for command injection if the agent does not properly escape the inputs.
[EXTERNAL_DOWNLOADS]: The skill relies on npx @openant-ai/cli@latest to fetch and execute its core CLI tool. Although the package belongs to the verified author openant-ai, the use of the @latest tag means the skill's behavior could change without notice if the package is updated.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it ingests untrusted text from users.
Ingestion points: User-provided inputs for the --title and --description flags in the SKILL.md file.
Boundary markers: Not present. There are no instructions or delimiters provided to signal that the model should ignore instructions embedded within user content.
Capability inventory: The skill has Bash access to manage tasks and interact with crypto wallets.
Sanitization: Not present. No escaping or validation logic is specified for the interpolated user content.

create-task