The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user or agent to install an external package globally using npm install -g sequential-thinking-cli or pnpm add -g sequential-thinking-cli. This package is a third-party binary retrieved from the public NPM registry without version pinning or a verified source origin.
[COMMAND_EXECUTION]: The skill's primary operation involves executing the sthink command via a system shell. This command-line interface is provided by the unverified external package mentioned above, meaning the agent's reasoning process and output generation depend on the integrity of this external binary.
[INDIRECT_PROMPT_INJECTION]: The skill defines a surface where external or user-generated reasoning content is processed through CLI commands.
Ingestion points: Data enters the process via the --content argument in the sthink step command as described in SKILL.md.
Boundary markers: The instructions use specific command flags (--content) to delimit the reasoning text from the command execution logic.
Capability inventory: The skill utilizes CLI execution (sthink), session state management, and file system write operations through the replay command's export functionality.
Sanitization: There is no explicit documentation regarding the sanitization or escaping of shell-sensitive characters within the reasoning content passed to the CLI, which may present a command injection risk depending on the implementation of the external tool.

sequential-thinking