The Agent Skills Directory

[SAFE]: No security issues were detected. The skill implements its functionality through a transparent shell script and well-documented API endpoints.
[COMMAND_EXECUTION]: The skill includes a bash script (scripts/crawlora.sh) that acts as a wrapper for curl. This script is safely implemented using set -euo pipefail for error handling and uses --data-urlencode to ensure that parameters passed to the Crawlora API are correctly sanitized and do not allow for command injection.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to api.crawlora.net. This is the official API domain of the skill's author (crawlora-org) and is used for its intended purpose of retrieving normalized food and restaurant data.
[CREDENTIALS_UNSAFE]: The skill requires an API key for access. It correctly advises the user to provide this key via an environment variable (CRAWLORA_API_KEY) and includes explicit warnings against hardcoding or committing the credential, adhering to standard security practices.
[PROMPT_INJECTION]: The skill ingests third-party data such as Yelp and OpenTable reviews, which represents a potential surface for indirect prompt injection.
Ingestion points: External data is ingested into the agent's context through the output of the scripts/crawlora.sh script.
Boundary markers: No specific boundary delimiters are defined in the instructions, although the API returns structured JSON which helps demarcate content.
Capability inventory: The skill possesses the ability to execute shell commands and perform network operations via curl.
Sanitization: The helper script ensures parameters are URL-encoded, but the agent should be aware that the content of retrieved reviews is untrusted.

restaurant-food-delivery-research