The Agent Skills Directory

[COMMAND_EXECUTION]: The orchestrator script orchestrate_edge_pipeline.py executes various research stages using subprocess.run. It follows security best practices by passing command arguments as a list and explicitly avoiding shell=True, which prevents shell-injection vulnerabilities.
[COMMAND_EXECUTION]: Execution is strictly limited to a predefined set of internal scripts mapped to specific pipeline stages (auto_detect, hints, concepts, drafts, review, export). The script validates that the requested stage exists in its hardcoded mapping before attempting execution.
[PROMPT_INJECTION]: The skill is designed to handle external data (hints and tickets) which are subsequently processed by other AI agents, creating an indirect prompt injection surface.
Ingestion points: Untrusted data enters the pipeline through files specified by the --llm-ideas-file and --tickets-dir arguments in scripts/orchestrate_edge_pipeline.py.
Boundary markers: The orchestrator does not implement specific boundary markers or "ignore" instructions when passing this data to downstream scripts.
Capability inventory: The orchestrator has the capability to execute subprocesses and write manifest files to the local file system.
Sanitization: No content-level sanitization is performed on the ingested YAML or JSON data before it is passed through the pipeline stages.

edge-pipeline-orchestrator