The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill is designed to search for and extract sensitive credentials, including Supabase anon keys, service role keys, and JWT secrets, as part of the Phase 2 'Key Extraction' process. These findings are stored in local tracking files (.sb-pentest-context.json) rather than being exfiltrated to an external service.
[INDIRECT_PROMPT_INJECTION]: The skill ingests a user-provided target URL that is used to perform subsequent scans. This constitutes an ingestion point for potentially untrusted data that could attempt to influence the agent's behavior during the audit.
Ingestion points: User-provided target URL (SKILL.md).
Boundary markers: The skill does not define specific boundary markers for the URL input.
Capability inventory: Network requests via curl, local file system writes for logging and evidence, and invocation of sub-skills.
Sanitization: No explicit sanitization or validation of the target URL is defined in the orchestration instructions.
[DYNAMIC_EXECUTION]: The orchestrator automatically generates a shell script (curl-commands.sh) containing the commands used during the audit. This is a low-risk script generation intended to provide the user with reproducible evidence of identified security gaps.

supabase-pentest