The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines patterns for processing external data (e.g., payment inputs, report requests, and order data) in background tasks across rules/jobs-task-queue.md, rules/temporal-activities.md, and rules/temporal-workflows.md. This creates an indirect prompt injection surface where malicious data could influence downstream agent actions if results are processed without sanitization.
Ingestion points: Task handlers like send_email, process_payment, and OrderWorkflow ingest untrusted data from external sources.
Boundary markers: The code examples do not demonstrate the use of delimiters or instructions to ignore embedded commands within the processed data.
Capability inventory: The skill patterns include significant capabilities such as performing network operations (requests, httpx), executing external activities, and managing state transitions based on external input.
Sanitization: The provided snippets do not include explicit input validation or sanitization logic to mitigate risks from untrusted task payloads.

async-jobs