The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/replay_dlq.py uses subprocess.run to execute commands constructed by substituting placeholders in a template with data from an external JSON file. While the default template uses the benign echo command, the script is specifically designed to be adapted for executing arbitrary CLI tools (such as temporal, aws, or curl), which could lead to command injection if a user or agent provides a malicious input file and the command template is insecurely configured.
[INDIRECT_PROMPT_INJECTION]: The skill provides scripts for processing external data (scripts/check_workflow_idempotency.py and scripts/replay_dlq.py), which presents a surface for indirect prompt injection. If an agent is tasked with analyzing or running these scripts on untrusted files, its behavior or outputs could be manipulated by content within the JSON data.
Ingestion points: scripts/check_workflow_idempotency.py (via stdin or file argument) and scripts/replay_dlq.py (via file argument).
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the script logic.
Capability inventory: The scripts/replay_dlq.py script has the capability to execute subprocesses.
Sanitization: The scripts parse JSON but do not perform validation or sanitization of the values before they are used in command construction or logic checks.

software-workflow-automation