The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes vendor-specific logistics data and documentation URLs. However, it implements robust boundary markers and safety instructions to mitigate risks. Specifically, SKILL.md and references/integration-checklist.md require the agent to label unknowns, use source-backed facts only, and avoid performing live actions (moving money, PII, or identity data) without explicit human approval. The ingestion points are limited to vendor markdown files and a source YAML map.
[COMMAND_EXECUTION]: A script scripts/list-vendors.mjs is included to provide a deterministic list of vendors. This script uses standard Node.js file system operations (readFileSync) to read a local JSON file (../../../data/processed/vendors.json). It does not execute arbitrary shell commands or take untrusted user input that could lead to command injection.
[DATA_EXFILTRATION]: While the skill involves network URLs for documentation, these are all static references to official logistics providers (e.g., Aramex, Bosta, SPL). The skill instructions explicitly warn against printing or committing secrets and emphasize protecting customer PII like addresses and phone numbers.
[REMOTE_CODE_EXECUTION]: No patterns of downloading and executing remote scripts (e.g., curl | bash) or dynamic code execution were detected. The skill focuses on providing integration guidance rather than automated script execution.

mena-logistics