The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from files generated during workflow execution, creating a surface for indirect instructions to influence the agent.
Ingestion points: The scripts/collect_results.py script reads multiple Markdown files from the .workflow/<slug>/results/ directory to create a summary checklist.
Boundary markers: The SKILL.md instructions include guidelines for explicit integration and resolving conflicts, but there are no technical delimiters or "ignore instructions" headers applied to the content read from files.
Capability inventory: The skill enables the agent to write to the local filesystem, execute bundled Python scripts, and potentially spawn subagents or enter sustained execution modes.
Sanitization: The Python scripts do not perform sanitization, filtering, or escaping of the content read from the result files before presenting it to the agent.
[COMMAND_EXECUTION]: The skill utilizes several local Python scripts to automate workflow management tasks.
Evidence: SKILL.md provides commands for the agent to run new_workflow.py, collect_results.py, and verify_workflow.py. These scripts are part of the skill package and use standard library modules for file and directory operations.

codex-dynamic-workflows