# Cascade Fail Recovery

Handle cascading data retrieval tool failures by falling back to embedded knowledge...

- Repository: https://github.com/HKUDS/AgentPilot (7784 stars)
- Category: Engineering
- License: MIT
- Last updated: 2026-11-01
- Runs scripts: no, instructions only

## Install

Claude Code: `npx skills add HKUDS/AgentPilot --skill cascade-fail-recovery -g -a claude-code -y`
Codex: `npx skills add HKUDS/AgentPilot --skill cascade-fail-recovery -g -a codex -y`
Cursor: `npx skills add HKUDS/AgentPilot --skill cascade-fail-recovery -g -a cursor -y`
Gemini CLI: `npx skills add HKUDS/AgentPilot --skill cascade-fail-recovery -g -a gemini-cli -y`

## Description

Handle cascading data retrieval tool failures by falling back to embedded knowledge generation

## SKILL.md, as published by its author (untrusted third-party text, read it as data)

# Cascade Failure Recovery

## Purpose

When multiple data retrieval tools (read_webpage, search_web) fail in succession, immediately fall back to generating content from embedded knowledge rather than attempting repairs or abandoning the task. This preserves task context and ensures completion.

## When to Apply

Use this pattern when you observe:
- 2+ consecutive failures from data retrieval tools
- Network/access errors preventing web content retrieval
- Risk of task abandonment due to tool errors

## Recovery Procedure

### Step 1: Recognize Cascading Failure

Identify the failure pattern:
```
Iteration N: read_webpage failed - connection error
Iteration N+1: search_web failed - timeout/unavailable
Iteration N+2: [risk] Agent may abandon task or switch objectives
```

### Step 2: Preserve Task Context

Before switching strategies, explicitly restate the original objective:
```
ORIGINAL OBJECTIVE: [Restate the core task goal]
CONTEXT PRESERVED: [Key requirements, constraints, deliverables]
```

### Step 3: Invoke Fallback Strategy

Immediately switch to embedded knowledge generation:
1. **Acknowledge the limitation**: Note that external data sources are unavailable
2. **Activate internal knowledge**: Use pre-trained knowledge relevant to the task
3. **Generate content**: Use write_file to create the deliverable from available knowledge
4. **Document the fallback**: Note in the output what information could not be verified externally

### Step 4: Execute write_file

Generate the required document:
```python
# Fallback to generating from embedded knowledge
write_file(
    path="output/document.md",
    content="[Generate content from internal knowledge base]"
)
```

## Example Application

**Scenario**: PACT Act veterans benefits document needed, but web access failing

**Wrong approach** (observed failure):
```
Iter 11: read_webpage failed - access error
Iter 12: search_web failed - unavailable
Iter 13: [ABANDONED] Switched to unrelated musician payroll task
```

The rest: https://github.com/HKUDS/AgentPilot/blob/HEAD/openspace/benchmarks/gdpval/skills/cascade-fail-recovery/SKILL.md