How do I install the Micro Task Workflow skill?

Run npx skillmds@latest add majiayu000/micro-task-workflow-2 in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Micro Task Workflow safe to use?

This skill has not completed SkillMD's automated safety review yet. Capability flags: docs only. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Micro Task Workflow?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Micro Task Workflow free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Micro Task Workflow?

majiayu000 (@majiayu000) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Micro Task Workflow by majiayu000 — Agent Skill

Name: Micro Task Workflow
Rating: 4.2 (5 reviews)
Author: majiayu000

File contents

micro-task-workflow Skill

Micro-task constraints and context budget management.

For complete workflow, see .claude/agents/orchestrator.md

When to Use This Skill

Understanding micro-task constraints
Escape hatch protocol when hitting context limits
Planning dependencies between tasks

The Problem: Context Exhaustion

Agents running out of context mid-task produce broken, incomplete work:

❌ OLD: Issue #100 "Implement Feature X"
├── Read 8 files to understand (~20% context)
├── Plan changes (~10% context)
├── Edit 4 files (~30% context)
├── Debug issues (~20% context)
├── Run tests (~15% context)
└── CONTEXT EXHAUSTED at 95% - work incomplete, uncommitted

The Solution: 50% Budget

✅ NEW: Issue #100 "Implement Feature X"
├── Micro-Task 100.1: Setup + config changes (45% context) ✓ committed
├── Micro-Task 100.2: Core implementation (45% context) ✓ committed
├── Micro-Task 100.3: Tests + documentation (45% context) ✓ committed
└── All work committed, PR ready

Micro-Task Constraints

Constraint	Limit	Rationale
File reads	≤ 5 files	Minimize exploration
File edits	≤ 3 files	Single logical change
Tool calls	≤ 80 total	~50% of context capacity
Commits	1-2	Checkpoint + final
Scope	Single concern	Complete in one session

Context Budget Breakdown

Phase	Budget	Purpose
Startup overhead	~20%	Load AGENTS.md, read issue, read source files
Productive work	~50%	Actual implementation
Safety margin	~30%	Unexpected complexity, debugging

Escape Hatch Protocol

At 60% context usage (or ~60 tool calls):

Commit current progress (even if incomplete):

git add -A && git commit -m "WIP(#100): partial progress"
git push

Write session state to .claude/session-state.md
Sync beads:
```
bd sync
```
Exit cleanly - do not continue until fresh session

Dependency Types

Serial micro-tasks (must run sequentially):

Same file modified by both (merge conflicts)
Output of one is input to another
Database schema changes before queries
API endpoint before frontend integration

Parallel micro-tasks (can run simultaneously):

Different files entirely
Same issue, independent concerns (e.g., tests vs docs)
Different issues with no shared files
Frontend and backend on different endpoints

Parallelization Rules

Scenario	Parallel?	Reason
Different issues, different files	Yes	No conflicts
Same issue, independent files	Yes	No conflicts
Same file modified	No	Merge conflicts
Sequential dependency	No	Output needed
Database migration + queries	No	Schema dependency

Worktree Strategy

Single worktree per issue (recommended):

All micro-tasks for issue #51 run in worktree 51-feature-name
Sequential micro-tasks commit to same branch
Squash to single commit before PR

Squash before PR:

git rebase -i main           # Squash all commits
git push --force-with-lease

---name: micro-task-workflow-2description: Micro-task development patterns with 50% context budget. Use for task decomposition, context management, escape hatch protocols, and orchestration patterns.---# micro-task-workflow Skill**Micro-task constraints and context budget management.**> For complete workflow, see `.claude/agents/orchestrator.md`## When to Use This Skill- Understanding micro-task constraints- Escape hatch protocol when hitting context limits- Planning dependencies between tasks---## The Problem: Context ExhaustionAgents running out of context mid-task produce broken, incomplete work:```❌ OLD: Issue #100 "Implement Feature X"├── Read 8 files to understand (~20% context)├── Plan changes (~10% context)├── Edit 4 files (~30% context)├── Debug issues (~20% context)├── Run tests (~15% context)└── CONTEXT EXHAUSTED at 95% - work incomplete, uncommitted```## The Solution: 50% Budget```✅ NEW: Issue #100 "Implement Feature X"├── Micro-Task 100.1: Setup + config changes (45% context) ✓ committed├── Micro-Task 100.2: Core implementation (45% context) ✓ committed├── Micro-Task 100.3: Tests + documentation (45% context) ✓ committed└── All work committed, PR ready```---## Micro-Task Constraints| Constraint | Limit | Rationale ||------------|-------|-----------|| File reads | ≤ 5 files | Minimize exploration || File edits | ≤ 3 files | Single logical change || Tool calls | ≤ 80 total | ~50% of context capacity || Commits | 1-2 | Checkpoint + final || Scope | Single concern | Complete in one session |### Context Budget Breakdown| Phase | Budget | Purpose ||-------|--------|---------|| Startup overhead | ~20% | Load AGENTS.md, read issue, read source files || Productive work | ~50% | Actual implementation || Safety margin | ~30% | Unexpected complexity, debugging |---## Escape Hatch Protocol**At 60% context usage (or ~60 tool calls):**1. **Commit current progress** (even if incomplete): ```bash git add -A && git commit -m "WIP(#100): partial progress" git push ```2. **Write session state** to `.claude/session-state.md`3. **Sync beads:** ```bash bd sync ```4. **Exit cleanly** - do not continue until fresh session---## Dependency Types**Serial micro-tasks** (must run sequentially):- Same file modified by both (merge conflicts)- Output of one is input to another- Database schema changes before queries- API endpoint before frontend integration**Parallel micro-tasks** (can run simultaneously):- Different files entirely- Same issue, independent concerns (e.g., tests vs docs)- Different issues with no shared files- Frontend and backend on different endpoints### Parallelization Rules| Scenario | Parallel? | Reason ||----------|-----------|--------|| Different issues, different files | Yes | No conflicts || Same issue, independent files | Yes | No conflicts || Same file modified | No | Merge conflicts || Sequential dependency | No | Output needed || Database migration + queries | No | Schema dependency |---## Worktree Strategy**Single worktree per issue** (recommended):- All micro-tasks for issue #51 run in worktree `51-feature-name`- Sequential micro-tasks commit to same branch- Squash to single commit before PR**Squash before PR:**```bashgit rebase -i main # Squash all commitsgit push --force-with-lease```

Frequently asked questions

Micro Task Workflow