Back to list
sharpner

delegation

by sharpner

1🍴 0📅 Jan 9, 2026

SKILL.md


name: delegation description: Use when deciding whether to delegate work to subagents. Lists available agents with triggers and capabilities. Delegate aggressively.

Subagent Delegation

Principle: Delegate aggressively. Fresh subagents work better than overloaded context.

When to Delegate

SituationAction
Complex code review→ Spawn code-reviewer agent
Security concerns→ Spawn security reviewer
UI/Mobile changes→ Spawn mobile/design reviewer
Large exploration (10+ files)→ Spawn Explore agent
Multiple independent tasks→ Spawn parallel agents

Available Subagents

Review Agents

AgentTriggerFocus
pr-review-toolkit:code-reviewerComplex PRs (>10 files)Code quality, patterns
pr-review-toolkit:silent-failure-hunterError handling changesSilent failures, catch blocks
pr-review-toolkit:type-design-analyzerNew types introducedType invariants, encapsulation
loose-ends-hunterRefactoring, deletionsDead code, orphaned imports

Utility Agents

AgentWhen
ExploreCodebase exploration, pattern search
PlanImplementation strategy, architecture
gemini-explorer[OPTIONAL] Large-context analysis (1M tokens)

Custom Project Agents

Define in .claude/agents/:

  • security-reviewer.md - OWASP Top 10 checks
  • mobile-reviewer.md - Touch targets, responsive
  • design-reviewer.md - Design system compliance

Parallel Dispatch Pattern

When multiple independent issues exist:

# Launch ALL in same message block = PARALLEL
Task(subagent_type="code-reviewer", prompt="Review PR #123 for code quality")
Task(subagent_type="security-reviewer", prompt="Review PR #123 for security")
Task(subagent_type="mobile-reviewer", prompt="Review PR #123 for mobile")

Rules for parallel dispatch:

  • One agent per independent problem domain
  • No shared state between investigations
  • Each agent gets complete context
  • Review and integrate results

Gemini vs Claude Subagents

FeatureGemini (via script)Claude Subagent
Cost~$0.02/review~$0.10/review
Context1M tokens~200K tokens
Speed30-40 sec60-90 sec

Use Gemini for: Diff-based reviews (security, mobile, design) Use Claude for: Complex analysis (loose-ends, user-journey)

Gemini Scripts (Optional)

# PR Review with verdict
./scripts/gemini-review.sh 236

# Gemini impersonates Claude agent
./scripts/gemini-subagent-impersonation.sh 236 security-reviewer

# Research topic
./scripts/gemini-research.sh "performance optimization"

Output Format

When delegating:

**Delegation:**
- Task: [what needs review]
- Agent: [which subagent]
- Reason: [why this agent]
- Dispatched: ✅ / Running in background

Score

Total Score

50/100

Based on repository quality metrics

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

Reviews

💬

Reviews coming soon