The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the ordercli binary from a third-party source not listed as a trusted vendor.
Evidence: Metadata defines installation via Homebrew (steipete/tap/ordercli) and Go (github.com/steipete/ordercli/cmd/ordercli@latest).
[DATA_EXFILTRATION]: The skill uses commands that access sensitive files and application data, specifically browser cookies and session profiles used for authentication.
Evidence: Use of commands like ordercli foodora cookies chrome --profile "Default" and references to accessing browser profile data in "$HOME/Library/Application Support/ordercli/browser-profile".
[CREDENTIALS_UNSAFE]: The instructions promote the handling of sensitive authentication tokens and passwords via command-line arguments and standard input.
Evidence: Mention of DELIVEROO_BEARER_TOKEN, DELIVEROO_COOKIE, and the --password-stdin flag for logins.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external food delivery APIs that could potentially contain malicious instructions embedded in order history or item descriptions.
Ingestion points: ordercli foodora history, ordercli foodora orders (SKILL.md).
Boundary markers: None identified in the provided instructions.
Capability inventory: Execution of local binaries (ordercli) and potential file system writes for configuration storage.
Sanitization: No sanitization logic is described for processing external API responses.

ordercli