ds-implement
REQUIRED Phase 3 of /ds workflow. Enforces output-first verification at each step via delegation.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
القائمة
REQUIRED Phase 3 of /ds workflow. Enforces output-first verification at each step via delegation.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
REQUIRED Phase 5 of the /ds workflow (final). Enforces reproducibility demonstration and user acceptance before completion claims.
REQUIRED Phase 4 of the /ds workflow to review methodology, data quality, and statistical validity. Provides structured review checklists and confidence scoring.
REQUIRED Phase 2 of /ds workflow. Profiles data and creates analysis task breakdown. Requires docs/SPEC.md from Phase 1.
This skill should be used when the user asks to "start a data science project", "brainstorm analysis", "plan a data analysis", or wants to clarify analysis requirements. REQUIRED Phase 1 of /ds workflow. Uses Socratic questioning to clarify goals, data sources, and constraints.
| name | ds-implement |
| description | REQUIRED Phase 3 of /ds workflow. Enforces output-first verification at each step via delegation. |
YOU MUST NOT WRITE ANALYSIS CODE. This is not negotiable.
You orchestrate. Subagents analyze. STOP if you're about to write Python/R code.
docs/*.md..py/.ipynb files in main chat.If you're about to write analysis code directly, STOP and spawn a Task agent instead.
EVERY CODE STEP MUST PRODUCE VISIBLE OUTPUT. Before moving to the next step, you (via subagents) MUST:
docs/LEARNINGS.md.cat docs/PLAN.md # View task sequence
For each task in docs/PLAN.md, spawn a Task agent with this protocol: Plaintext
Task(subagent_type="general-purpose", prompt=""" Implement [TASK] following output-first protocol. Ref: ~/.config/opencode/skills/ds-implement/references/verification-patterns.md
Required steps:
Every step must be logged with:
Input: Shape/State.
Operation: Exact transformation.
Output: Resulting shape, null counts, and sample data.
Verification: Why the result is considered "correct".
Red Flags - STOP Immediately
"I'll check results at the end" (Silent error compounding).
"This transform is too simple to print" (Hidden logic bugs).
"I know the merge worked" (Assumptions lead to data loss).
No Pause Between Tasks
After completing task N, IMMEDIATELY start task N+1. You MUST NOT pause. Do not ask "Should I proceed?" or "Would you like a summary?". Keep moving until the plan is complete or you are blocked. Phase CompleteAfter all tasks in docs/PLAN.md are verified, IMMEDIATELY invoke: /skill ds-review