메뉴
Execution behavior for one saved automation job running in a fresh agent conversation.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
Create, inspect, edit, archive, delete, run, and search scheduled or executable automation assets.
Teaches agents how to author Markdown that renders cleanly into DOCX via write_docx, with heading hierarchy (up to 6 levels), tables, lists, blockquotes, code blocks, citations, embedded images, SVG diagrams (rendered to PNG first), and document structure patterns for reports, proposals, and technical documents.
Teaches agents how to write syntactically correct Mermaid diagrams that render without errors across all supported diagram types. Covers quoting rules, common syntax pitfalls, diagram-type-specific syntax (flowchart, sequence, class, ER, Gantt, state, pie, mindmap), subgraphs, styling with classDef and theme variables, KDCube brand color integration, renderer usage patterns, readability tuning, and complexity limits.
Teaches agents how to generate Markdown, HTML (with embedded SVG), and Mermaid content that renders beautifully to multi-page PDF via write_pdf, with proper page breaks, compact professional layouts, brand and domain-adaptive color schemes, multi-column support for scientific papers, magazine-style editorial documents, and iterative render-review-fix workflow.
Teaches agents how to author slide-structured HTML that renders professionally to PPTX via write_pptx, with proper sizing, styling, brand and domain-adaptive color schemes, SVG diagram integration, citation handling, deck narrative flow, and content budgets for business presentations, technical decks, and executive briefings.
Teaches agents how to hand-craft self-contained SVG diagrams that render correctly in browsers, <img> tags, and Playwright PDF/PNG. Covers containment math, viewBox sizing, arrow construction, text hierarchy, KDCube brand color tokens, semantic zone fills, badges, the PNG inspection loop, write_png parameters, multi-diagram workflows, and common failure modes with fixes.
| name | job |
| id | job |
| description | Execution behavior for one saved automation job running in a fresh agent conversation. |
| version | 1.0.0 |
| category | automation |
| tags | ["automations","execution","journal"] |
| when_to_use | ["The agent (you) is now running a saved automation as a manual or scheduled job.","The agent (you) must record substantial execution progress or final outcome.","The agent (you) needs read-only access to the current automation or linked automations."] |
| author | kdcube |
| created | "2026-05-03T00:00:00.000Z" |
| namespace | automation |
You are executing one saved automation in a fresh job conversation. Do not manage the user's automation list in this mode. Do not create, edit, delete, or link automation definitions. Do not create or update user memory.
Use automation_job.get_current_automation to load the automation you are running. It may include
explicitly linked automation definitions for bounded context. Do not pass a automation id;
the bundle injects the current automation id through runtime context.
Use job_memory.search_memo only when durable user context would materially
change how this job should run.
Use automation_job.search_automation_executions to inspect prior outcomes for the current
automation, or explicitly linked automations when relevant. Do not pass automation ids; the bundle
injects the current automation id through runtime context.
Use automation_job.update_execution_journal when progress, result data, error
details, or produced artifacts must not be lost. Do not pass automation id or
execution id; the bundle injects both through runtime context. Keep journal
entries compact and user-facing. Avoid dumping raw logs unless they are
important evidence.
Call automation_job.* tools directly as normal ReAct tool calls. Do not call them
from inside exec_tools.execute_code_python generated Python; automation_job is not
available as a Python global in the isolated exec runtime.
When you produce files, include artifact metadata in the journal:
logical_path for React fi: referencesfilenamemime_typedescriptionhosted_uri when the file is already deliverable to the userFor news/search/report automations, use web/search tools as needed and preserve the requested output format, for example HTML. The final journal should mention the main findings and produced artifact metadata.
At the end of the job, update the execution journal one last time with
success, failed, or cancelled, then provide a concise final answer.