opendfm/xcientist/code-planning
Claude Code and agent skill by opendfm/xcientist
Trust tier 3
- community
Source check pending
Checking that the SKILL.md file is still in the repository.
Build an ordered executable code plan for controlled experiment enablement
Install
npx vskill@latest install opendfm/xcientist/code-planningSKILL.md
Code Planning
Mission
Translate the idea into an ordered executable code plan that supports:
- All-components science conditions where every canonical idea component is enabled
- Component-disabled science conditions using the same runner interface with one or more components disabled
CRITICAL: Code must be SELF-CONTAINED in `project/`
- All code must be COMPLETE and INDEPENDENT in
project/. - The code must NOT depend on
repos/for core functionality. - Everything needed to run experiments must be in
project/. - Repos are for REFERENCE ONLY - implement actual experiment code in
project/.
Protocol
- Write the managed artifact
code.planatagent_reports/code/plan/latest.jsonwith ordered stages, project target paths, expected outputs, and verification commands. - Map each idea.json component to code implementation node(s).
- For each component node, document the component-disable mechanism (how to turn that component off through the unified runner).
- Sequence stages so each worker can finish with a clean, runnable
project/
state before the next stage starts.
- Generate the managed
code.blueprintartifact before delegating implementation. - End the plan with a mandatory
final_integration_smokenode that uses the real prepared dataset and the real API/model path when required. - The real benchmark data from
dataset_candidate/MUST be used in experiments, not synthetic data. - Keep planner outputs, worker reports, reviewer reports, hook reports, phase summaries, and controlled artifacts under the structured
agent_reports/code/layout. - Merge worker outputs into
agent_reports/code/summary.md, optionalagent_reports/code/usage.md,agent_reports/code/artifacts/integration_readiness.json, and the reviewer-approved phase reportagent_reports/code/phase.json.
Report Requirements
The runtime owns reports. Do not write report files with tools. Each step keeps:
agent_reports/code/worker/<step>/attempts/NNN.jsonandlatest.jsonagent_reports/code/review/<step>/attempts/NNN.jsonandlatest.jsonagent_reports/code/hook/<step>/attempts/NNN.jsonandlatest.json
Phase-level reports:
agent_reports/code/plan/latest.json- Ordered step planagent_reports/code/plan/planner_report.json- Planner summaryagent_reports/code/summary.md- Human-readable code statusagent_reports/code/usage.md- How to run experiments (optional)agent_reports/code/artifacts/integration_readiness.json- Readiness statusagent_reports/code/phase.json- Final phase verdict
Hard Rules
- Do not claim a condition is enabled unless a runnable command exists.
- Do not rely on parallel code execution; the runtime executes the ordered
stages list and gates each stage before moving on.
final_integration_smokemust not be satisfied by imports, mocks, or dry-run-only evidence.- Every idea.json component must have a corresponding component-disable mechanism in the plan.
- The plan must include runnable entrypoints/configuration for all-components and component-disabled science conditions.
- Every step must finish through the worker/review/hook prefinish gate.
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingsrc/agents/experiment_agent/skills/code-planning/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/opendfm/xcientist/code-planning)