Agent skill

escalate

Use when stuck in circles and want to hand off to another agent (Codex CLI) with full context

Stars 163
Forks 31

Install this agent skill to your Project

npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/data/escalate

SKILL.md

Escalate

Generate a handoff prompt for another agent when you're stuck and need fresh perspective.

When to Use

  • You've tried multiple approaches and keep hitting the same walls
  • You're going in circles and want a fresh perspective
  • You need to context-switch but want to preserve progress for Codex CLI or another agent

Workflow

/escalate → Choose detail level → Choose handoff mode → Analyze session → Generate prompt → pbcopy

Step 1: Ask Detail Level

Use AskUserQuestion to ask:

How much context for the handoff?

1. Concise (~200-400 words) - Problem, key blockers, clear ask
2. Detailed (~500-800 words) - Above plus files, errors, code snippets

Step 2: Ask Handoff Mode

Use AskUserQuestion to ask:

What should the next agent do?

1. Continue - Pick up where I left off and solve this
2. Fresh perspective - Ignore my approaches, try something new
3. Validate first - Confirm my understanding, then propose alternatives
4. Other - [let user specify custom instruction]

Step 3: Analyze the Session

Review the conversation to extract:

Section What to Capture
Problem The core objective - what you're trying to accomplish
Approaches Tried Each attempt and why it didn't work
Current Hypothesis What you think might be the root cause
Relevant Files Key file paths that are central to the issue
Error Messages Specific errors encountered (exact text)
Constraints Things that must be preserved or avoided

For concise mode: Focus on Problem, Approaches Tried, and one key blocker.

For detailed mode: Include all sections with code snippets and full error messages.

Step 4: Generate the Prompt

Template

markdown
# Problem

[Clear statement of what you're trying to accomplish]

# Approaches Tried

- **Approach 1:** [what you tried]
  - Result: [what happened]
  - Why it failed: [analysis]

- **Approach 2:** [what you tried]
  - Result: [what happened]
  - Why it failed: [analysis]

[Continue for each significant approach]

# Current Hypothesis

[What you think might be the root cause or blocker]

# Relevant Context

**Files:**
- `path/to/file.ts` - [why it's relevant]
- `path/to/other.ts` - [why it's relevant]

**Errors:**

[Exact error messages]


**Constraints:**
- [Things that must be preserved]
- [Approaches to avoid and why]

# Your Task

[Based on handoff mode:]

- Continue: "Pick up where I left off. The context above shows what's been tried. Find a path forward and solve this."
- Fresh perspective: "Ignore the approaches above - they haven't worked. Look at this problem fresh and propose a different strategy."
- Validate first: "First, confirm whether my understanding of the problem is correct. Then propose an alternative approach."
- Other: [User's custom instruction]

Step 5: Copy to Clipboard

bash
echo "[generated prompt]" | pbcopy

Use a heredoc for multi-line content:

bash
pbcopy << 'EOF'
[generated prompt content]
EOF

Step 6: Confirm

Tell the user:

Handoff prompt copied to clipboard. Paste into Codex CLI to continue.

Quality Checklist

  • Problem statement is clear and self-contained
  • Approaches include WHY they failed, not just WHAT was tried
  • Hypothesis is specific, not vague
  • File paths are absolute or repo-relative
  • Error messages are exact, not paraphrased
  • Constraints call out landmines to avoid
  • Handoff instruction matches selected mode

Common Mistakes

Mistake Fix
Vague problem statement Be specific about the goal and success criteria
Listing approaches without failure analysis Always explain WHY each approach didn't work
Missing context Include enough that the next agent doesn't need to ask clarifying questions
Too much detail in concise mode Ruthlessly cut to ~300 words for concise
Forgetting constraints Warn about approaches that look promising but have been ruled out

Example Output (Concise)

markdown
# Problem

Trying to get Jest tests to run in a WordPress plugin. Tests pass locally but fail in CI with "Cannot find module '@wordpress/scripts'" errors.

# Approaches Tried

- **Verified node_modules:** Confirmed package is installed, same versions local and CI
- **Cleared CI cache:** Rebuilt from scratch, same error
- **Checked path resolution:** Added moduleNameMapper in jest.config.js, no change

# Current Hypothesis

CI environment might have different NODE_PATH or module resolution behavior. The @wordpress/scripts package uses a nested dependency structure.

# Your Task

Pick up where I left off. The context above shows what's been tried. Find a path forward and solve this.

Didn't find tool you were looking for?

Be as detailed as possible for better results