The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is instructed to run several local commands to gather environment and configuration data, specifically git remote -v, tea login ls, and tea issues create --help. These are standard discovery operations for the intended functionality.
[DYNAMIC_EXECUTION]: The skill generates a shell command string for the user to execute: tea issues create ... --title "<title>" --description "$(</tmp/tea-issue-<slug>.md)". The use of double quotes for the --title argument is unsafe because POSIX-compliant shells will perform command substitution (e.g., $(...) or `...`) and variable expansion inside double quotes. If the drafted title contains such characters, they will be executed by the user's shell upon pasting.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection as it is instructed to 'explore the codebase' and use 'user-provided instructions' to draft the issue content.
Ingestion points: Codebase files, conversation history, and user instructions (SKILL.md Instruction 1).
Boundary markers: None. The skill does not use delimiters or instructions to ignore embedded commands in the data it processes.
Capability inventory: Accesses git configuration, CLI login states, and writes files to /tmp (SKILL.md Instruction 4 and Requirement 2).
Sanitization: The skill lacks instructions to sanitize or escape shell metacharacters in the generated title before presenting the command to the user.

draft-tea-command-for-issue-create