スキル一覧に戻る
ramtinJ95

egenskriven

by ramtinJ95

Manually managed dotfiles

0🍴 0📅 2026年1月18日
GitHubで見るManusで実行

SKILL.md


name: egenskriven description: Local-first kanban task manager for AI agent workflows. Use when managing tasks, tracking work, creating issues, or when user mentions tasks, kanban, boards, backlog, or egenskriven.

Overview

EgenSkriven is a local-first kanban task manager with CLI and web UI, designed for AI agents from the ground up. It replaces ephemeral todo systems (like TodoWrite) with persistent, structured task tracking.

Quick Start Commands

# Get project status summary
egenskriven context --json

# Get recommended next task
egenskriven suggest --json

# List actionable (unblocked) tasks
egenskriven list --ready --json

# Create a new task
egenskriven add "Task title" --type feature --priority medium

Essential CRUD Operations

CommandDescription
add "title"Create task (flags: --type, --priority, --column)
listList tasks (flags: --ready, --column, --type)
show <ref>Display task details
move <ref> <column>Move task between columns
update <ref>Modify task properties
delete <ref>Remove task (--force skips confirmation)

Task Reference Formats

Tasks can be referenced by:

  • Full ID: abc123def456
  • ID prefix: abc (minimum unique prefix)
  • Display ID: WRK-123 (board prefix + sequence)
  • Title substring: "dark mode" (case-insensitive match)

Columns (Workflow States)

ColumnDescription
backlogIdeas and future work
todoReady to start
in_progressCurrently being worked on
need_inputBlocked, waiting for human input
reviewAwaiting review
doneCompleted

JSON Output

All commands support --json for machine-readable output:

# Get specific fields only (reduces tokens)
egenskriven list --json --fields id,title,column

# Full task details
egenskriven show <ref> --json

Exit Codes

CodeMeaning
0Success
1General error
2Invalid arguments
3Task not found
4Ambiguous reference
5Validation error

When to Use EgenSkriven

  • Multi-step work spanning multiple exchanges
  • Work that might be interrupted
  • Features with dependencies
  • Bugs that need tracking

Not needed for: Simple questions, one-off commands, trivial changes.

Tool Integrations

Before using the collaborative workflow (blocking tasks, resume flow), initialize the tool integration for your AI coding tool:

OpenCode

egenskriven init --opencode

This creates .opencode/tool/egenskriven-session.ts. When working on a task, call the egenskriven-session tool to get your session ID:

Agent: [calls egenskriven-session tool]
Response: { session_id: "abc-123", link_command: "egenskriven session link <task> --tool opencode --ref abc-123" }
Agent: [runs link command to associate session with task]

Claude Code

egenskriven init --claude-code

This creates:

  • .claude/hooks/egenskriven-session.sh - Hook script that runs on SessionStart
  • .claude/settings.json - Hook configuration (merged with existing settings)

After the hook runs, your session ID is available as $CLAUDE_SESSION_ID. Link your session with:

egenskriven session link <task-ref> --tool claude-code --ref $CLAUDE_SESSION_ID

Codex CLI

egenskriven init --codex

This creates .codex/get-session-id.sh. Get your session ID with:

SESSION_ID=$(.codex/get-session-id.sh)
egenskriven session link <task-ref> --tool codex --ref $SESSION_ID

All Integrations

Generate all tool integrations at once:

egenskriven init --all
egenskriven init --all --force  # Overwrite existing files

Human-AI Collaborative Workflow

EgenSkriven supports a collaborative workflow where AI agents can request human input and resume work once they receive a response. This enables seamless back-and-forth communication between humans and AI coding assistants.

Workflow Overview

Agent starts work on task
        │
        ▼
Agent encounters decision point
        │
        ▼
Agent blocks task: egenskriven block <task> "question"
        │
        ▼
Human sees blocked task in UI or CLI
        │
        ▼
Human adds response: egenskriven comment <task> "response"
        │
        ▼
Resume (manual, command, or auto)
        │
        ▼
Agent continues with full context

Setting Up Tool Integration

Before using the collaborative workflow, set up the integration for your AI tool:

OpenCode

egenskriven init --opencode

Creates .opencode/tool/egenskriven-session.ts. Call the egenskriven-session tool to get your session ID.

Claude Code

egenskriven init --claude-code

Creates hooks that set $CLAUDE_SESSION_ID automatically on session start.

Codex CLI

egenskriven init --codex

Creates .codex/get-session-id.sh helper script.

Linking Your Session

Before blocking a task, link your session:

# OpenCode: Call the egenskriven-session tool first, then:
egenskriven session link <task-ref> --tool opencode --ref <session-id>

# Claude Code:
egenskriven session link <task-ref> --tool claude-code --ref $CLAUDE_SESSION_ID

# Codex:
SESSION_ID=$(.codex/get-session-id.sh)
egenskriven session link <task-ref> --tool codex --ref $SESSION_ID

Blocking a Task

When you need human input:

# Block with a question (atomic operation)
egenskriven block <task-ref> "What authentication approach should I use?"

# For longer questions, use stdin
echo "I need to decide between several approaches..." | egenskriven block <task-ref> --stdin

This will:

  1. Move the task to the need_input column
  2. Add your question as a comment
  3. Preserve your session for later resume

Adding Comments

# Add a response comment
egenskriven comment <task-ref> "Use JWT with refresh tokens"

# Add with explicit author
egenskriven comment <task-ref> "Approved" --author "jane"

# Add from stdin for longer responses
cat response.txt | egenskriven comment <task-ref> --stdin

Viewing Comments

# List all comments
egenskriven comments <task-ref>

# As JSON
egenskriven comments <task-ref> --json

# Only recent comments
egenskriven comments <task-ref> --since "2026-01-07T10:00:00Z"

Listing Blocked Tasks

# List all tasks needing input
egenskriven list --need-input

# As JSON
egenskriven list --need-input --json

Resume Modes

Configure per-board:

ModeBehavior
manualPrint command for user to copy
commandUser runs egenskriven resume --exec
autoAuto-resume on @agent comment
# Set resume mode
egenskriven board update <board> --resume-mode auto

Resuming Work

Depending on the board's resume mode:

# Print the resume command (manual mode)
egenskriven resume <task-ref>

# Execute the resume directly (command mode)
egenskriven resume <task-ref> --exec

# Use minimal prompt (fewer tokens)
egenskriven resume <task-ref> --exec --minimal

For auto mode, add a comment with @agent and the session will resume automatically.

Example: Complete Workflow

# 1. Agent links session
egenskriven session link WRK-42 --tool opencode --ref abc-123

# 2. Agent works, then gets blocked
egenskriven block WRK-42 "Should I implement REST or GraphQL first?"

# 3. Human responds
egenskriven comment WRK-42 "Start with REST, we'll add GraphQL later"

# 4. Human resumes agent
egenskriven resume WRK-42 --exec

# Agent continues with full context...

Tips

  • Always link your session BEFORE blocking a task
  • Be specific in your blocking questions
  • Use the --json flag for scripting and automation
  • Check comments periodically if working on multiple tasks
  • egenskriven-workflows - Workflow modes (strict/light/minimal) and agent behaviors
  • egenskriven-advanced - Epics, dependencies, sub-tasks, batch operations

スコア

総合スコア

50/100

リポジトリの品質指標に基づく評価

SKILL.md

SKILL.mdファイルが含まれている

+20
LICENSE

ライセンスが設定されている

0/10
説明文

100文字以上の説明がある

0/10
人気

GitHub Stars 100以上

0/15
最近の活動

3ヶ月以内に更新がある

0/10
フォーク

10回以上フォークされている

0/5
Issue管理

オープンIssueが50未満

+5
言語

プログラミング言語が設定されている

+5
タグ

1つ以上のタグが設定されている

0/5

レビュー

💬

レビュー機能は近日公開予定です