The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill instructs the agent to dynamically generate and run Python scripts using the execute_code tool to orchestrate file processing and web data extraction.
[INDIRECT_PROMPT_INJECTION]: The workflow involves ingesting untrusted data from URLs and codebase files which are then processed by subagents.
Ingestion points: web_search, web_extract, and read_file (SKILL.md).
Boundary markers: Absent.
Capability inventory: execute_code, delegate_task, terminal, and patch (SKILL.md).
Sanitization: Absent, though logical verification via 'adversarial convergence' is recommended.
[PERSISTENCE]: The skill documents methods for maintaining long-running workflows across sessions using cronjob and background terminal processes.
[COMMAND_EXECUTION]: The orchestration patterns include the use of shell commands and filesystem modifications via terminal and patch tools.

dynamic-workflow