Back to list
0xrdan

retry

by 0xrdan

Intelligent model orchestration for Claude Code - routes queries to optimal Claude model (Haiku/Sonnet/Opus) based on complexity. It also includes many more features. If this project is working well for you and would like to support me, just help spread the word. Thanks!

22🍴 3📅 Jan 18, 2026

SKILL.md


name: retry description: Retry the last query with an escalated model user_invokable: true

Retry Skill

Retry the last query with a more capable model when the initial attempt failed or was insufficient.

What This Does

When a query routed to a cheaper model (Haiku or Sonnet) produces unsatisfactory results, errors, or timeouts, use /retry to:

  1. Re-route to the next tier up (Haiku -> Sonnet -> Opus)
  2. Preserve the original query context
  3. Give the more capable model a chance to succeed

Usage

/retry              # Escalate to next tier
/retry deep         # Force escalation to Opus
/retry standard     # Force escalation to Sonnet

When to Use

  • Timeout or error: Query failed due to complexity
  • Incomplete answer: Model didn't fully address the question
  • Wrong approach: Model misunderstood the task
  • Need more depth: Initial answer was too superficial

How It Works

  1. Read the last routing decision from session state (~/.claude/router-session.json)
  2. Determine the appropriate escalation:
    • fast (Haiku) -> standard (Sonnet)
    • standard (Sonnet) -> deep (Opus)
    • deep (Opus) -> Already at max, suggest different approach
  3. Re-execute the last query with the escalated model
  4. Update session state with the new route

Instructions

When this skill is invoked:

  1. Check for explicit model - Did user specify deep/opus or standard/sonnet?
    • YES: Use that model. DO NOT auto-escalate. Honor the explicit choice.
    • NO: Proceed to escalation logic
  2. Read session state from ~/.claude/router-session.json
  3. Determine escalation (only if no explicit model):
    • From fast: Escalate to standard
    • From standard: Escalate to deep
    • From deep: Already at max, suggest different approach
  4. Inform the user of the escalation
  5. Spawn the appropriate subagent using Task tool

CRITICAL: If user specifies a model (/retry deep, /retry standard), use exactly that model. No exceptions.

Example

User ran a complex refactoring query that was routed to Haiku. Haiku produced incomplete results.

User: /retry
Assistant: Escalating from Haiku to Sonnet...
           Re-running your refactoring query with more capable model.
           [Spawns standard-executor with the original query]

Notes

  • This skill reads the session state, which persists for 30 minutes
  • If no previous query exists, inform the user
  • Consider the failure reason when suggesting the escalation level

Score

Total Score

75/100

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

+10
説明文

100文字以上の説明がある

+10
人気

GitHub Stars 100以上

0/15
最近の活動

1ヶ月以内に更新

+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

+5

Reviews

💬

Reviews coming soon