The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill manages sensitive Temporal and AWS configuration. It proactively warns against hardcoding API keys and provides templates for secure secret management via environment variables or secret stores, ensuring credentials are not leaked in logs or shell history.
[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: Worker dependencies are sourced from the official 'temporalio' organization on public registries. The skill provides specific installation instructions and versioning guidance for the serverless SDK modules, ensuring the integrity of the build process.
[DYNAMIC_EXECUTION]: The workflow involves runtime assembly of worker code and infrastructure templates. The agent is guided through standard compilation and packaging steps (e.g., Go cross-compilation, Python wheel isolation) to produce legitimate Lambda deployment packages.
[PRIVILEGE_ESCALATION]: CloudFormation templates are used to provision IAM roles with scoped permissions (lambda:InvokeFunction, lambda:GetFunction). These roles utilize sts:ExternalId to mitigate cross-account confused deputy risks, and the skill clarifies the distinction between execution and invocation roles.
[INDIRECT_PROMPT_INJECTION]: The skill processes data entering the agent context via CLI outputs (e.g., aws lambda list-functions, temporal worker deployment list). Ingestion points are the shell outputs in setup.md and iam.md. While explicit boundary markers and sanitization are absent, the skill implements a strict capability inventory (aws/temporal subprocess calls and file writes) and instructs the agent to verify all actions against state read back from the provider to mitigate accidental command execution.

temporal-serverless