Навык forge
Implement tasks from a change. Use when the user wants to start or continue implementation of an existing change.
- Звёзды на GitHub
- 13
- Файлы
- 1
- Последний коммит
- 23 июн. 2026 г.
- Лицензия
- MIT
- Разрешённые инструменты
- —
Что делает навык forge?
forge — это Agent Skill. Implement tasks from a change. Use when the user wants to start or continue implementation of an existing change.
Как установить навык forge
Каждая команда копирует папку навыка в каталог навыков клиента. Перезапустите клиент или начните новую сессию, чтобы загрузить его.Установка навыка forge в Плагин
/plugin marketplace add devnomad-byte/techneering /plugin install tn@techneering
Установка навыка forge в Claude Code
~/.claude/skills/forge/SKILL.md
$ npx degit devnomad-byte/techneering/skills/forge ~/.claude/skills/forgeУстановка навыка forge в Claude Code (project)
.claude/skills/forge/SKILL.md
$ npx degit devnomad-byte/techneering/skills/forge .claude/skills/forgeУстановка навыка forge в Codex
~/.agents/skills/forge/SKILL.md
$ npx degit devnomad-byte/techneering/skills/forge ~/.agents/skills/forgeУстановка навыка forge в Gemini CLI
~/.gemini/skills/forge/SKILL.md
$ npx degit devnomad-byte/techneering/skills/forge ~/.gemini/skills/forgeУстановка навыка forge в Claude.ai
npx degit devnomad-byte/techneering/skills/forge forge && zip -r forge.zip forge # claude.ai → Customize → Skills → + → Create skill → Upload a skill → forge.zip
SKILL.md
Открыть на GitHubForge a Change
Implementation entry point. Load change artifacts, then chain to tn:blueprint for task breakdown and tn:assemble for execution.
Core Principle
READ ALL CONTEXT BEFORE WRITING ANY CODE.
Load every artifact. Understand the full picture. Then implement.
Steps
Step 1: Select the Change
If a name is provided, use it. Otherwise:
- Infer from conversation context if the user mentioned a change
- Auto-select if only one active change exists in
techneering/changes/ - If ambiguous, list all active changes with progress info and ask the user to select
Always announce: "Using change: "
Step 2: Load Context
Read all artifacts from the change directory:
techneering/changes/<name>/proposal.mdtechneering/changes/<name>/specs/**/*.mdtechneering/changes/<name>/design.mdtechneering/changes/<name>/tasks.md
Also read techneering/config.yaml for project context.
Step 3: Show Current Progress
Parse tasks.md for checkboxes: - [ ] (pending) vs - [x] (complete).
Display:
- Change name and schema
- Progress: "N/M tasks complete"
- Remaining tasks overview
Step 4: Create Implementation Plan (tn:blueprint)
Invoke the tn:blueprint skill to break down the tasks into bite-sized implementation steps:
- Exact file paths
- Complete code for each step
- Verification commands with expected output
- No placeholders
Step 5: Execute with Assemble (tn:assemble)
Invoke the tn:assemble skill to execute the plan:
- Each task gets a fresh subagent
- Two-stage review per task: spec compliance → code quality
- Independent tasks can run in parallel (max 3 concurrent)
- Follow
tn:redgreenfor each task (red-green-refactor) - Use
tn:diagnoseif issues arise
Step 6: Track Progress
After each task completes:
- Mark the corresponding checkbox in tasks.md:
- [ ]→- [x] - Report progress: "Task N/M complete"
Step 7: On Completion
Display:
## Implementation Complete
**Change:** <change-name>
**Progress:** N/N tasks complete
### Completed This Session
- [x] Task 1
- [x] Task 2
...
All tasks complete! Proceed to tn:gate to verify.
Step 8: On Pause (Issue Encountered)
Display:
## Implementation Paused
**Change:** <change-name>
**Progress:** N/M tasks complete
### Issue Encountered
<description>
**Options:**
1. <option 1>
2. <option 2>
What would you like to do?
Red Flags
| Thought | Reality |
|---|---|
| "I remember the requirements, no need to read files" | Lost context is the #1 cause of bugs |
| "Write code first, read spec later" | Spec is the source of truth — read it first |
| "This task is too simple for a blueprint" | All tasks go through blueprint → assemble |
| "Skip review to speed up" | Review is quality assurance — not optional |
Quick Reference
| Dimension | Key point |
|---|---|
| Role | Implementation entry point, chains blueprint → assemble |
| Input | Change name (optional, can auto-detect) |
| Internal chain | forge → blueprint → assemble (cannot be skipped) |
| Progress tracking | Mark tasks.md checkbox after each task |
| Path placeholder | Always use techneering/changes/<name>/ |
Next Step
forge is a composite skill that chains internally: tn:blueprint → tn:assemble, then exits to tn:gate. Users do not need to trigger the intermediate steps manually.
When done:
- All tasks complete → auto-invoke
tn:gate(then show menu) - Issue encountered → invoke
tn:diagnose - User wants to pause → save progress, wait for next session
- User requests stop → end the current flow
Guardrails
- Always read all context files before starting
- Keep code changes minimal and scoped to each task
- Update task checkbox immediately after completing each task
- Pause on errors, blockers, or unclear requirements — don't guess
- If implementation reveals design issues, pause and suggest artifact updates
- Unified path placeholder: always use
techneering/changes/<name>/
Файлы в этом навыке
1Вопросы о навыке forge
Как установить навык forge в Claude Code?
Выполните npx degit devnomad-byte/techneering/skills/forge ~/.claude/skills/forge. Claude Code загружает навыки из ~/.claude/skills (все проекты) или .claude/skills (один проект) и использует навык, когда его описание соответствует задаче.
Какие клиенты могут использовать навык forge?
Любой клиент, который читает формат Agent Skills: Claude Code, Codex, Gemini CLI и claude.ai (загрузить как ZIP). Скрипты в навыке выполняются только там, где клиент разрешает выполнение кода.