The Agent Skills Directory

[PERSISTENCE]: The skill modifies the shell configuration file ~/.zshrc to store environment variables and command aliases, ensuring that configured providers remain active across terminal sessions.
[CREDENTIALS_UNSAFE]: The skill stores API keys for multiple providers (DeepSeek, Z.ai, DashScope, MiniMax, OpenRouter) in plain text within the user's ~/.zshrc file. While it attempts to secure these with chmod 600 and umask 077 during modification, the keys remain stored in the filesystem.
[COMMAND_EXECUTION]: The skill executes several internal shell scripts (detect-mode.sh, check-status.sh, write-alias.sh, verify-providers.sh) to perform its core functions of status checking and file modification.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to external API endpoints (e.g., openrouter.ai, deepseek.com, z.ai) using curl to verify that the provided API keys are valid.
[DYNAMIC_EXECUTION]: The skill uses eval within verify-providers.sh to load keys from the .zshrc file and pipes curl output to python3 -c for parsing and validating model identifiers.
[INDIRECT_PROMPT_INJECTION]: The maintainer-only 'Update' mode uses sub-agents to browse external documentation, creating a potential surface for indirect prompt injection if those websites contain adversarial content.

provider-switch