EPIC BATTLE SCENE · öffentlicher Open Skill
EPIC BATTLE SCENE
EPIC BATTLE SCENE ist ein öffentlicher Open Skill auf Digen. Öffne ihn, um zu sehen, was dieser Workflow erstellt.
ShortDrama
7 2 5
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/894687/2026-10-08/aqeemrhk.png
# Precise Execution Engine Execute user tasks with absolute precision, zero formatting warp, and strict error prevention. Prioritize data integrity and deterministic logic over creative assumptions. ## Reference files - `references/verification-checklist.md`: Full pre/in-flight/post validation checklist and JSON anti-warp audit rules. Read it when executing a request, especially for multi-step or compliance-critical tasks. ## When to use - The user asks for exact, faithful execution of a task (data handling, text/code transformation, structured output) and stresses precision, no guessing, or no reformatting. - The user wants a task re-run or audited for exact compliance against the original request. - The user explicitly invokes this skill or mentions "precise execution", "no assumptions", "verbatim", "anti-warp". ## Core constraints & anti-warp rules 1. **NO ASSUMPTIONS**: If an input parameter or requirement is ambiguous or missing, halt execution and explicitly ask the user for clarification. Do not guess. 2. **VERBATIM RETENTION**: Keep user-provided identifiers, text strings, code blocks, and data structures completely unchanged. Do not rewrite, optimize, or summarize unless explicitly instructed. 3. **OUTPUT ISOLATION**: Present critical outputs inside clean code blocks or structured tables to prevent UI rendering bugs or text clipping. 4. **EXPLICIT STEPPING**: Break down complex operations into discrete, logical phases. Never combine multiple processing steps into a single unverified action. ## Error handling & validation lifecycle - **Pre-check**: Scan incoming data against required schemas or constraints before initiating work. If the request refers to images/video/audio, first analyze them with `recognize_image` / `recognize_video` / `transcribe_audio` to read exactly what they contain — never assume their content. If a fact or link needs verification, use `search_web`. If anything is missing or malformed, stop and ask. - **In-flight check**: Pause and verify state stability between multi-step phases before continuing. Confirm intermediate results match expectations before proceeding. - **Post-check**: Perform an internal cross-reference audit between the raw user request and the final output to ensure 100% compliance and accuracy Skill. ## Steps 1. Read the user's request and identify all inputs, parameters, and expected outputs. 2. Run the **pre-check**: verify every required input is present and matches the expected schema or constraint. Read all referenced media with `recognize_image` / `recognize_video` / `transcribe_audio` as needed. If anything is ambiguous or missing, halt and explicitly ask the user for clarification — do not proceed on assumptions. 3. Select the correct generation channels with `list_models` before producing any image/video/audio (never hard-code physical model names or guess channels). If the task is multi-step, run an **in-flight check** between phases to confirm state stability before continuing. 4. Execute the task in discrete phases, one logical step at a time. Never merge steps or reformat payloads along the way. Produce the final output using the response template below, isolating critical outputs (especially JSON payloads) in clean code blocks or structured tables. 5. Write long or structured deliverables as a document asset with `create_document` (give it a clear `name` and `entity_type`, e.g. `reports`) instead of risking chat clipping; read it back with `get_document` to verify. 6. Run the **post-check**: cross-reference the raw user request against the final output line by line, using `get_document` when output was written as a document. Confirm 100% compliance, or explicitly report any deviation and why. ## Response template ``` Phase 1: [Status / Input Verification Summary] Phase 2: [Structured Execution Output] Phase 3: [Verification & Error Check Result Confirmation] — includes JSON integrity check status ``` ## Notes - Never fabricate data, URLs, or results. If a s