# Develop

> Develop phase entry; routes task execution, state updates, KB sync, and plan close-out through references/assets/scripts.

- Skill: `evidentloop/develop` (Agent Skill, multi-file: 6 files)
- Install (CLI): `npx skillmds@latest add evidentloop/develop`
- Raw SKILL.md: https://api.skillmd.com/api/skills/evidentloop/develop/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: evidentloop (https://skillmd.com/u/evidentloop)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/evidentloop/develop

---


# Develop (Entry)

> Core philosophy: No verification evidence, no completion — "it should be fine" is not evidence.

## When to activate

- Entering the implementation phase (`workflow` / `light_iterate` / `quick_fix`; bare `~go` auto-routes to `exec_plan` when an active plan exists).
- Need to execute the task list, update state, and converge the delivery result.

## Execution skeleton

1. Read the active plan tasks (`tasks.md` or light `plan.md`).
2. Extract pending tasks and execute them in numbered order.
3. Update task markers after each step (`[ ] -> [x] / [-] / [!]`).
4. Sync KB files and conservative preference / feedback records.
5. Set completed plans' `plan.md` lifecycle metadata to `ready_to_archive` and keep them in `plan/`.
6. Render the matching result template.

## Resource navigation

- Long rules: `references/develop-rules.md`
- Shared writing standards: `../references/shared-writing-dna.md` (apply to all output)
- Output contract: `../references/output-contract.md` (required sections, density gradient, symbol discipline, desensitization)
- Output templates: `assets/*.md`
- Task extraction script: `scripts/extract_pending_tasks.py`

## Deterministic logic first

Use the script when task extraction must be auditable:

```bash
python3 skills/en/skills/sopify/develop/scripts/extract_pending_tasks.py \
  --tasks-file .sopify/plan/<plan>/tasks.md
```

The script returns JSON with pending tasks, status counts, and execution order.

## Boundaries

- This skill executes and closes out work; it does not redefine the plan structure.
- `ready_to_archive` is `plan.md` metadata only; archival happens only when explicit `~go finalize` delegates to `sopify_writer`.
- Rollback remains an explicit user action and must keep a traceable record.