Skip to content
Agentmart
Submit a tool

three-shifts

Cyclic shift execution engine.

Skill by profbernardoj

IndexedMIT★ 110
Write a review

Chalk Labs

Need help deploying three-shifts?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/profbernardoj-three-shifts-f47e94

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Cyclic shift execution engine. Plans tasks 3x daily (6 AM, 2 PM, 10 PM), decomposes them into granular steps, then executes via 15-minute cron cycles. Each cycle reads state files, picks the next step, executes it, writes results back. Errors are logged and skipped — never fatal. Planning uses Claude 4.6; execution uses GLM-5.

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/three-shifts

What's inside

Skill name
three-shifts
Description
Cyclic shift execution engine. Plans tasks 3x daily (6 AM, 2 PM, 10 PM), decomposes them into granular steps, then executes via 15-minute cron cycles. Each cycle reads state files, picks the next step, executes it, writes results back. Errors are logged and skipped — never fatal. Planning uses Claude 4.6; execution uses GLM-5.
License
MIT
Instructions
457 lines in SKILL.md · view on GitHub

Files (6)

  • SKILL.md
  • config.md · references/
  • context.md · templates/
  • handoff.md · templates/
  • state.json · templates/
  • tasks.md · templates/

SKILL.md (first lines)

# Three Shifts v2 — Cyclic Execution Engine

**Architecture:** Plan once → decompose → execute in 15-minute cycles → each cycle does ONE step → state persists in files between runs.

**Why v2:** V1 ran everything in one long session. One timeout killed the whole shift. V2 is a state machine — errors get logged, skipped, and retried. 32 cycles per shift, each learning from the last.

---

## File Structure

All state lives in `workspace/shifts/`:

```
shifts/
├── tasks.md        # Current shift's step-by-step plan (the state machine)
├── context.md      # Stack, constraints, known pitfalls, learned lessons
├── handoff.md      # What the previous shift left behind
├── state.json      # Machine-readable shift metadata
└── history/        # Archived completed shifts
    └── YYYY-MM-DD-{shift}.md
```

### tasks.md Format

```markdown
# Afternoon Shift — 2026-02-22
Approved by: user | Approved at: 2:15 PM CST
Shift window: 2:00 PM – 10:00 PM CST
Cycles remaining: ~30

## Task 1: API migration plan [P1]
- [x] Step 1.1: Read current proxy-router config and session economics — DONE (cycle 1)
- [ ] Step 1.2: Research mor-[REDACTED] paid tier pricing and limits
- [!] Step 1.3: Compare cost of paid tier vs local-only — BLOCKED: need pricing from step 1.2
- [ ] Step 1.4: Draft transition recommendation in memory/projects/example/

## Task 2: Commit workspace changes [P1]
- [ ] Step 2.1: Run git status, identify logical commit groups
- [ ] Step 2.2: Stage and commit new skills (three-shifts, relationships)
- [ ] Step 2.3: Stage and commit memory reorganization files
- [ ] Step 2.4: Stage and commit scripts and config changes
- [ ] Step 2.5: Push to remote

## Summary
Total steps: 9 | Done: 1 | Blocked: 1 | Remaining: 7
Last cycle: 2026-02-22 14:15 CST — completed step 1.1
Next target: Step 1.2
```

**Step markers:**
- `[ ]` — Not started (available for next cycle)
- `[x]` — Completed (with cycle number and brief result)
- `[!]` — Blocked (with reason — will be retried after dependencies clear)
- `[~]` — In progress (claimed by current cycle, prevents double-execution)
- `[-]` — Skipped (manually removed or deprioritized)

### context.md Format

```markdown
# Shift Context — Living Document

## Stack & Environment
- macOS arm64, OpenClaw 2026.2.21-2
- Git repos: ~/.openclaw/workspace
- 1Password for secrets (never store raw keys)
- Primary messaging channel for user comms

## Constraints
- Never force push or delete branches
- trash > rm (recoverable beats gone)
- Space Brave searches by 2+ seconds (rate limit)
- MiniMax-M2.5: DO NOT USE (broken streaming)
- Llama 3.3 70B: DO NOT USE

## Known Pitfalls (learned from experience)
- API-specific gotchas go here
- Check messaging connectivity if deliveries fail
- Check provider billing cooldown policies
- Always check git status before committing — avoid partial commits

## Rules for This Shift
- Night shift: no external comms, no financial txns, no destructive ops
- If blocked on user input, skip and note in handoff
- Max 1 web search per cycle (Brave rate limit)
```

### state.json Format

```json
{
  "shift": "afternoon",
  "date": "2026-02-22",
  "status": "executing",
  "startedAt": "2026-02-22T20:00:00Z",
  "approvedAt": "2026-02-22T20:15:00Z",
  "approvedBy": "user",
  "planModel": "your-heavy-model (e.g. Claude, GPT-4)",
  "execModel": "your-light-model (e.g. GLM-5, GPT-4o-mini)",
  "totalSteps": 9,
  "completed": 1,
  "blocked": 1,
  "skipped": 0,
  "cyclesRun": 1,
  "lastCycleAt": "2026-02-22T20:15:00Z",
  "nightAutoApproved": false,
  "carryoverFromShift": null
}
```

---

## Shifts

| Shift | Plan Cron | Cycle Crons | Window | Character |
|-------|-----------|-------------|--------|-----------|
| Morning | 6:00 AM | 6:15 AM – 1:45 PM (every 15 min) | 6 AM – 2 PM | Comms, decisions, user-facing |
| Afternoon | 2:00 PM | 2:15 PM – 9:45 PM (every 15 min) | 2 PM – 10 PM | Deep work, building, coding |
| Night | 10:00 PM | 10:15 PM – 5:45 AM (every 15 min) | 10 PM – 6 AM | Autonomous, research, maintenance |

From profbernardoj/everclaw-community-branches, MIT.

More from profbernardoj/everclaw-community-branches

Signals

Stars
110
Forks
20
Open issues
1
Last commit (this folder)
May 28, 2026 (5 months ago)
Last release
May 15, 2026 (5 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

Same skill elsewhere

Copies of this SKILL.md published by others; search shows only one of them.

More from profbernardoj

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get three-shifts running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying three-shifts?

CallGet help
Cite this page

three-shifts skill. Agent Mart. https://agentmart.dev/skills/profbernardoj-three-shifts-f47e94 (last updated 2026-10-24)