The Agent Skills Directory

[CREDENTIALS_UNSAFE] (MEDIUM): The skill explicitly requests the user's Cloudflare Global API Key and Account Email to perform its functions.
Evidence: Found in SKILL.md under 'Gather Credentials' and as required arguments in scripts/check_cloudflare_config.py and scripts/fix_ssl_mode.py.
Risk: Global API Keys provide full administrative access to the entire Cloudflare account, including billing and all zones. This is a high-privilege credential.
[DATA_EXFILTRATION] (LOW): The skill transmits the collected sensitive credentials (email and API key) to an external endpoint (api.cloudflare.com).
Evidence: Both Python scripts and the curl examples in SKILL.md use these credentials in HTTP headers (X-Auth-Email, X-Auth-Key) for network requests.
Mitigation: The destination is the legitimate Cloudflare API, but the handling of these secrets by an AI agent increases the risk of accidental exposure in chat logs or history.
[COMMAND_EXECUTION] (LOW): The skill relies on executing system commands (curl, dig, openssl) and Python scripts that perform network operations and state changes.
Evidence: SKILL.md contains multiple curl commands using PATCH and POST methods to modify SSL settings and purge caches.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill processes untrusted data from an external API which could theoretically contain malicious instructions.
Ingestion points: JSON responses from api.cloudflare.com are parsed in scripts/check_cloudflare_config.py and scripts/fix_ssl_mode.py.
Boundary markers: Absent; the agent is instructed to "Analyze evidence to determine root cause" based on the raw API output.
Capability inventory: The agent has the capability to modify account settings via PATCH requests.
Sanitization: The scripts use standard JSON parsing, but no content filtering is applied to string values returned by the API (e.g., DNS record names or comments) before they are presented to the agent.

cloudflare-troubleshooting