Mealie API skill

Interact with a self‑hosted Mealie instance (recipe manager & meal planner) via its REST API. Use for adding, updating, retrieving recipes, meal plans and generating shopping lists. Trigger when the user mentions their Mealie URL, wants to import a recipe, create a meal plan or fetch a shopping list.

Do not install or run this skill without first verifying and correcting the inconsistencies. Specifically: - Confirm the package actually includes scripts/mealie.sh (it is referenced in SKILL.md but not present in the file manifest). If it is missing, ask the author or reject the package. - Expect to provide MEALIE_URL and a sensitive MEALIE_TOKEN; the registry metadata should declare these — treat the token like a password, use a token with minimal scope, and rotate it regularly. - Ensure curl and jq are available on the system before use. Inspect any script contents before making them executable; do not run unknown shell scripts blindly. - Verify that the MEALIE_URL points to your intended self-hosted instance (to avoid sending tokens/data to an attacker-controlled host). - If you plan to let the agent invoke this autonomously, consider the blast radius of a leaked token and prefer short-lived tokens or a proxy that enforces limits. If the maintainer corrects the metadata (declare required env vars and required binaries) and includes the referenced script verbatim in the package, the skill would be coherent and appropriate for its stated purpose.

Type: OpenClaw Skill Name: mealie Version: 1.0.0 The skill bundle is classified as suspicious due to a significant vulnerability in `scripts/mealie.sh` (as described in `SKILL.md`). The `add-recipe` and `create-plan` commands use `curl --data @${1}`, which expects a file path. If the OpenClaw agent passes unsanitized user-provided input directly as this file path, it could lead to arbitrary local file reads (e.g., `/etc/passwd`) and exfiltration of their content to the user's configured Mealie instance. This represents a high-risk prompt injection vulnerability against the agent, allowing for potential data leakage, even though the skill itself does not exhibit explicit malicious intent like exfiltrating data to an unauthorized third party.