The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill search for and reads sensitive configuration files to extract credentials required for deployment checks.
Evidence: The instructions use Grep to scan .env*, .sentryclirc, and package.json for sensitive keys including SENTRY_AUTH_TOKEN, SUPABASE_SERVICE_ROLE_KEY, and LANGFUSE_SECRET_KEY.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from production logs and live web pages, creating a surface where malicious content could potentially influence the agent's logic.
Ingestion points: The skill reads external data via supabase:get_logs, npx langfuse-cli api traces list, and browsing the production site with playwright.
Boundary markers: There are no explicit instructions or delimiters used to separate untrusted external content from the agent's instructions.
Capability inventory: The skill utilizes tools like supabase:execute_sql, Grep, Glob, and npx commands.
Sanitization: No sanitization or filtering is performed on the ingested logs or web content before the agent processes them for verification reporting.
[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill downloads and executes command-line tools from external registries at runtime.
Evidence: Fetches and runs @playwright/cli@latest and langfuse-cli using npx. These are well-known tools from Microsoft and Langfuse utilized for their intended purpose.
[PRIVILEGE_ESCALATION]: The skill utilizes tools that provide high-privilege access to external database environments.
Evidence: Employs the supabase:execute_sql tool to run queries. While the current instructions focus on system statistics, the tool provides a mechanism for arbitrary SQL execution on the target database.

deploy-verify