The Agent Skills Directory

COMMAND_EXECUTION (HIGH): The skill parses user-provided URLs to extract a <task_id> and interpolates it into a shell command defined in ~/.bonfire/source.json. Step 1.1 and 1.2 describe executing these commands (e.g., node ... query.mjs get <task_id>). A maliciously crafted URL containing shell metacharacters (e.g., https://clickup.com/t/$(rm -rf $HOME)) could lead to arbitrary command execution if the parsing logic is not strictly sanitized.
EXTERNAL_DOWNLOADS (MEDIUM): The skill's 'Prerequisites' section directs users to install a GitHub CLI skill using npx skills add hairyf/skills. The source hairyf/skills is not a verified or trusted organization, representing a supply chain risk where the agent might download and execute untrusted third-party code.
REMOTE_CODE_EXECUTION (MEDIUM): The skill utilizes npx skills find to dynamically discover and confirm 'query methods' from the internet or local environment. These discovered commands are then saved to a persistent configuration file (~/.bonfire/source.json) and executed automatically in future sessions. This pattern allows for persistent execution of potentially malicious commands discovered via an unverified search process.

github-workflow