The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/campaign_admin.py executes git commands via subprocess.run to verify that output directories containing sensitive credentials (campaign API keys) are correctly ignored by version control. This is a defensive check to prevent accidental credential leakage.- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external store APIs which could theoretically contain malicious instructions.
Ingestion points: API responses (product titles, descriptions) processed in scripts/campaign_admin.py and stored in discovery.json.
Boundary markers: A mandatory plan review step (Phase 4) requiring explicit human operator approval via AskUserQuestion before any mutation occurs.
Capability inventory: Local file writes and scoped git command execution via subprocess.run.
Sanitization: The engine performs strict JSON schema validation, regex checks on slugs/prices, and decimal parsing on all data before use.- [SAFE]: The skill implements significant safety measures, including pacing API requests to avoid rate limits, refusing all HTTP redirects to mitigate SSRF risks, and pinning network requests to the authorized store origin and specific Cart API host to prevent cross-origin credential exfiltration.

next-create-campaign