All skills
simota avatar

/sherpa

@35ffd55
by shingo imotasimota/agent-skills85 stars
15

Guiding workflows by decomposing complex tasks (Epics) into Atomic Steps under 15 minutes each, with progress tracking and drift prevention. Use when complex decomposition is needed.

Use this Skill: https://skilld.dev/gh/simota/agent-skills/sherpa

This session only. Nothing lands on disk.

referenceprogress-tracking.md

≈1.2k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Progress Tracking

Purpose: Use this file when Sherpa needs dashboards, stalled detection, dependency analysis, retrospectives, or pace adjustment rules.

Contents

  • Epic dashboard
  • Stalled progress detection
  • Task tool integration
  • Dependency graph
  • Velocity tracking
  • Retrospectives
  • Adaptive pacing

Epic Dashboard

## Epic Dashboard: [Name]

**Started**: YYYY-MM-DD HH:MM
**Elapsed**: X hours Y minutes
**Status**: On Track | Behind | Blocked

### Progress
[========....................] 40% (4/10 steps)

### Velocity
- Completed: 4 steps in 45 min
- Average: 11 min/step
- Projected completion: 1h 15m remaining

### Steps
| # | Step | Size | Est | Actual | Status |
|---|------|------|-----|--------|--------|
| 1 | Define types | S | 10m | 8m | Done |
| 2 | API mock | M | 20m | 25m | Done |
| 3 | UI skeleton | M | 20m | 12m | Done |
| 4 | Form logic | S | 15m | - | In Progress |

### Commits
- `abc123` feat(payment): add payment types

### Blockers
- None currently

### Drift Log
- 14:23 - "Also fix footer" -> Added to backlog

Stalled Progress Detection

Condition Threshold Response
No progress > 30 min on one step prompt for help or further decomposition
Repeated attempts same step 3x suggest Scout investigation
Blocked external dependency switch to a valid parallel task
Overwhelmed user reports being stuck shrink the step

Stalled Response

## Progress Check

**Current Step**: [Step Name]
**Time on Step**: 35 minutes

**What might help?**
1. Break down further
2. Scout investigation
3. Pair with a specialist
4. Take a short break
5. Skip to a valid parallel task

Task Tool Integration

Sherpa persists progress with Claude Code task tools.

Lifecycle

TaskCreate (pending) -> TaskUpdate (in_progress) -> TaskUpdate (completed)

Metadata Example

TaskCreate({
  subject: "Define PaymentProps interface",
  description: "Create TypeScript interface for payment form props",
  activeForm: "Defining PaymentProps interface",
  metadata: {
    epicId: "payment-flow",
    stepNumber: 1,
    riskLevel: "green",
    estimatedMinutes: 10,
    agent: "Builder",
    parallelWith: [],
    blockedBy: []
  }
})

Resume Template

## Resuming Expedition

**Epic**: Payment Flow
**Last Session**: 2 hours ago
**Progress**: 4/10 steps completed

### Where We Left Off
- Step 5 was in progress
- Steps 6-10 pending

**Resume Options**:
1. Continue Step 5 (Recommended)
2. Review completed work
3. Re-plan the remaining work

Dependency Analysis

Type Symbol Meaning
Sequential -> A must finish before B
Parallel `
Blocking X external blocker
Optional ? can be skipped if time is tight
### Step Dependencies

| Step | Depends On | Blocks | Parallel With |
|------|------------|--------|---------------|
| 1 | - | 2, 3 | - |
| 2 | 1 | 4 | 3 |
| 3 | 1 | 4 | 2 |
| 4 | 2, 3 | 5 | - |

**Critical Path**: 1 -> 2 -> 4 -> 5
**Parallelizable**: 2 and 3

Velocity Tracking

### Velocity Analysis

**Session Stats**:
- Steps completed: 6
- Total time: 85 minutes
- Average: 14.2 min/step

**Calibration Factor**: 0.95x

| Step | Estimated | Actual | Delta |
|------|-----------|--------|-------|
| 1 | 10 min | 8 min | -20% |
| 2 | 15 min | 18 min | +20% |
| 3 | 10 min | 9 min | -10% |

Retrospectives

Full Retrospective

## Session Retrospective

**Date**: YYYY-MM-DD
**Duration**: Xh Ym
**Epic**: [Name]

### Progress Summary
- Started at: Step N
- Ended at: Step M
- Completed: X steps
- Commits: Y

### Metrics
| Metric | Value | Notes |
|--------|-------|-------|
| Steps/hour | X | Above/below average |
| Avg step time | Y min | On/off estimate |
| Drift incidents | Z | Managed/unmanaged |
| Blockers | N | Resolved/pending |

Quick Retro

## Quick Retro

**Completed**: Steps X-Y
**Blocked by**: Nothing / [blocker]
**Tomorrow**: Start Step Z
**Note**: [Key observation]

Adaptive Pacing

Mode When Step size Check-in frequency
Sprint fresh, deadline pressure normal (10-15 min) after each step
Cruise normal working mode normal every 2-3 steps
Recovery after blocker or break smaller (5-10 min) after each step
Wind-down end of session smallest, clean stops frequent
Increase step size when:
  - velocity > 1.2x estimate for 3+ steps
  - user requests a faster pace
  - work is simple and repetitive

Decrease step size when:
  - velocity < 0.8x estimate for 2+ steps
  - errors or drift are increasing
  - territory is unfamiliar
  - fatigue signals appear

Switch to Wind-down when:
  - session > 3 hours
  - user says "one more thing then done"
  - a natural stop point is near

Source: SKILL.md on GitHub

No alerts13d5 checks · Risk SAFE
  • Gen Agent Trust Hub13d

    The skill provides a robust framework for decomposing complex tasks into manageable atomic steps, featuring extensive protocols for risk management and focus protection. No security issues were identified.

  • Socket13d

    No alerts

  • Snyk13d

    Risk: LOW · No issues

  • Runlayer6mo

    1/11 files flagged

  • ZeroLeaks5mo

    Score: 93/100 · 2 sections analyzed

Signed by skilld at 35ffd55. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 days ago.

Activeupdated 2 weeks ago

README badge

README badge for simota/agent-skills/sherpa