AgentSkillsCN

worklog

添加带时间戳的工作日志条目，以追踪进度与决策过程。适用于记录工作内容、决策要点、潜在陷阱，以及智能体之间的交接事项。

SKILL.md
--- frontmatter
name: worklog
description: "Add timestamped work log entries to track progress and decisions. Use for documenting work, decisions, gotchas, and handoffs between agents."
model: claude-sonnet-4-20250514
allowed-tools: Read, Write, Edit, Glob

/worklog

Add structured JSON entries to track work, decisions, and learnings.

Usage

bash
/worklog yourbench YB-2 "Added login button to header"
/worklog yourbench YB-2 --decision "Using Clerk for auth"
/worklog yourbench YB-2 --gotcha "Token refresh needs cleanup"
/worklog coordinatr 003 --handoff code-reviewer "Ready for review"
/worklog yourbench YB-2 --state              # Show current state
/worklog yourbench YB-2 --migrate            # Migrate from WORKLOG.md

Directory Structure

code
ideas/yourbench/issues/YB-2-auth/
├── TASK.md
├── PLAN.md
└── worklog/
    ├── _state.json              # Current state (quick context load)
    ├── 001-phase-init.json      # Entry files
    └── 002-handoff-review.json

Entry Types

TypeFlagUse Case
Manual(default)General progress update
Decision--decisionDocument architectural choice
Gotcha--gotchaCapture lesson learned
Handoff--handoff TOAgent transition
Phase--phase NUMPhase completion
Blocker--blockerRecord impediment
Resolution--resolve IDResolve blocker

Execution Flow

1. Parse Arguments

code
/worklog PROJECT ISSUE_ID [--type] "message"

2. Locate Worklog Directory

bash
ideas/[project]/issues/[issue_id]-*/worklog/
mkdir -p [path] if missing

3. Get Next Sequence Number

bash
ls worklog/*.json | grep -v _state | wc -l
# Next = count + 1

4. Create Entry File

Filename: {sequence:03d}-{type}-{slug}.json

Required fields:

  • •$schema: "worklog-entry-v1"
  • •id: "ISSUE-SEQ"
  • •sequence: number
  • •timestamp: ISO 8601
  • •type: entry type
  • •author: { agent: string | null, human: string | null }
  • •summary: description

5. Update _state.json

After every entry:

  • •Update last_entry
  • •Update last_updated
  • •Increment entries_count
  • •Add to key_decisions if decision
  • •Update blockers if blocker/resolution

Viewing State

bash
/worklog yourbench YB-2 --state

Outputs:

code
Issue: YB-2 - Initialize Next.js project
Status: in_progress (Phase 3)
Progress: 5/5 phases complete
Key Decisions: ...
Blockers: none

Schema Reference

See references/schema.md for full JSON schema specification.

Best Practices

  1. •Be specific: Include enough context for future AI
  2. •Tag consistently: Use established tag taxonomy
  3. •Capture gotchas immediately: Don't wait until end
  4. •Handoff explicitly: Create handoff entry when switching agents
  5. •Update state: _state.json should always reflect current reality