The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the pinme command-line tool for authentication and uploading generated artifacts. These commands are integral to the skill's purpose of sharing static pages through the vendor's infrastructure.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from conversation histories, project metadata, and external files to create summaries and landing pages. Although it instructs the agent to escape text, this processing flow constitutes a surface for indirect injection.
Ingestion points: External workspace files, Codex conversation history, and user-provided project data used for share artifact generation.
Boundary markers: The skill includes instructions to "Escape user-provided text" but does not specify explicit structural prompt delimiters.
Capability inventory: Execution of shell commands via pinme upload using paths derived from user/project input.
Sanitization: Instructions explicitly mandate the removal of credentials, environment files (.env), private user data, and repository metadata (.git) before publication.

pinme-share