Escalation Governance

Table of Contents

• Overview
• The Iron Law
• When to Escalate
• When NOT to Escalate
• Decision Framework
• 1. Have I understood the problem?
• 2. Have I investigated systematically?
• 3. Is escalation the right solution?
• 4. Can I justify the trade-off?
• Escalation Protocol
• Common Rationalizations
• Agent Schema
• Orchestrator Authority
• Red Flags - STOP and Investigate
• Integration with Agent Workflow
• Quick Reference

Escalation Governance

Overview

Model escalation (haiku→sonnet→opus) trades speed/cost for reasoning capability. This trade-off must be justified.

Core principle: Escalation is for tasks that genuinely require deeper reasoning, not for "maybe a smarter model will figure it out."

The Iron Law

NO ESCALATION WITHOUT INVESTIGATION FIRST

Verification: Run the command with --help flag to verify availability.

Escalation is never a shortcut. If you haven't understood why the current model is insufficient, escalation is premature.

When to Escalate

Legitimate escalation triggers:

| Trigger | Description | Example |

|---------|-------------|---------|

| Genuine complexity | Task inherently requires nuanced judgment | Security policy trade-offs |

| Reasoning depth | Multiple inference steps with uncertainty | Architecture decisions |

| Novel patterns | No existing patterns apply | First-of-kind implementation |

| High stakes | Error cost justifies capability investment | Production deployment |

| Ambiguity resolution | Multiple valid interpretations need weighing | Spec clarification |

When NOT to Escalate

Illegitimate escalation triggers:

| Anti-Pattern | Why It's Wrong | What to Do Instead |

|--------------|----------------|---------------------|

| "Maybe smarter model will figure it out" | This is thrashing | Investigate root cause |

| Multiple failed attempts | Suggests wrong approach, not insufficient capability | Question your assumptions |

| Time pressure | Urgency doesn't change task complexity | Systematic investigation is faster |

| Uncertainty without investigation | You haven't tried to understand yet | Gather evidence first |

| "Just to be safe" | False safety - wastes resources | Assess actual complexity |

Decision Framework

Before escalating, answer these questions:

1. Have I understood the problem?

• [ ] Can I articulate why the current model is insufficient?
• [ ] Have I identified what specific reasoning capability is missing?
• [ ] Is this a capability gap or a knowledge gap?

If knowledge gap: Gather more information, don't escalate.

2. Have I investigated systematically?

• [ ] Did I read error messages/outputs carefully?
• [ ] Did I check for similar solved problems?
• [ ] Did I form and test a hypothesis?

If not investigated: Complete investigation first.

3. Is escalation the right solution?

• [ ] Would a different approach work at current model level?
• [ ] Is the task inherently complex, or am I making it complex?
• [ ] Would breaking the task into smaller pieces help?

If decomposable: Break down, don't escalate.

4. Can I justify the trade-off?

• [ ] What's the cost (latency, tokens, money) of escalation?
• [ ] What's the benefit (accuracy, safety, completeness)?
• [ ] Is the benefit proportional to the cost?

If not proportional: Don't escalate.

Escalation Protocol

When escalation IS justified:

• Document the reason - State why current model is insufficient
• Specify the scope - What specific subtask needs higher capability?
• Define success - How will you know the escalated task succeeded?
• Return promptly - Drop back to efficient model after reasoning task

Common Rationalizations

| Excuse | Reality |

|--------|---------|

| "This is complex" | Complex for whom? Have you tried? |

| "Better safe than sorry" | Safety theater wastes resources |

| "I tried and failed" | How many times? Did you investigate why? |

| "The user expects quality" | Quality comes from process, not model size |

| "Just this once" | Exceptions become habits |

| "Time is money" | Systematic approach is faster than thrashing |

Agent Schema

Agents can declare escalation hints in frontmatter:

model: haiku
escalation:
  to: sonnet                 # Suggested escalation target
  hints:                     # Advisory triggers (orchestrator may override)
    - security_sensitive     # Touches auth, secrets, permissions
    - ambiguous_input        # Multiple valid interpretations
    - novel_pattern          # No existing patterns apply
    - high_stakes            # Error would be costly

Verification: Run the command with --help flag to verify availability.

Key points:

• Hints are advisory, not mandatory
• Orchestrator has final authority
• Orchestrator can escalate without hints (broader context)
• Orchestrator can ignore hints (task is actually simple)

Orchestrator Authority

The orchestrator (typically Opus) makes final escalation decisions:

Can follow hints: When hint matches observed conditions

Can override to escalate: When context demands it (even without hints)

Can override to stay: When task is simpler than hints suggest

Can escalate beyond hint: Go to opus even if hint says sonnet

The orchestrator's judgment, informed by conversation context, supersedes static hints.

Red Flags - STOP and Investigate

If you catch yourself thinking:

• "Let me try with a better model"
• "This should be simple but isn't working"
• "I've tried everything" (but haven't investigated why)
• "The smarter model will know what to do"
• "I don't understand why this isn't working"

ALL of these mean: STOP. Investigate first.

Integration with Agent Workflow

**Verification:** Run the command with `--help` flag to verify availability.
Agent starts task at assigned model
├── Task succeeds → Complete
└── Task struggles →
    ├── Investigate systematically
    │   ├── Root cause found → Fix at current model
    │   └── Genuine capability gap → Escalate with justification
    └── Don't investigate → WRONG PATH
        └── "Maybe escalate?" → NO. Investigate first.

Verification: Run the command with --help flag to verify availability.

Quick Reference

| Situation | Action |

|-----------|--------|

| Task inherently requires nuanced reasoning | Escalate |

| Agent uncertain but hasn't investigated | Investigate first |

| Multiple attempts failed | Question approach, not model |

| Security/high-stakes decision | Escalate |

| "Maybe smarter model knows" | Never escalate on this basis |

| Hint fires, task is actually simple | Override, stay at current model |

| No hint fires, task is actually complex | Override, escalate |

Model Capability Notes

MCP Tool Search (Claude Code 2.1.7+): Haiku models do not support MCP tool search. If a workflow uses many MCP tools (descriptions exceeding 10% of context), those tools load upfront on haiku instead of being deferred. This can consume significant context. Consider escalating to sonnet for MCP-heavy workflows or ensure haiku agents use only native tools (Read, Write, Bash, etc.).

Claude.ai MCP Connectors (Claude Code 2.1.46+): Users with claude.ai connectors configured may have additional MCP tools auto-loaded, increasing the total tool description footprint. This makes it more likely that haiku agents will exceed the 10% tool search threshold. When escalation decisions involve MCP-heavy workflows, factor in claude.ai connector tool count via /mcp.

Effort Controls as Escalation Alternative (Opus 4.6 / Claude Code 2.1.32+): Opus 4.6 introduces adaptive thinking with effort levels (low, medium, high). The max level was removed in 2.1.72 for Opus 4.6, and high became the ceiling on that model. Claude Code 2.1.111 reintroduced max and added xhigh (between high and max) for Opus 4.7 only; on other models xhigh falls back to high. Symbols: ○ (low) ◐ (medium) ● (high) ◉ (xhigh) ★ (max). Use /effort (interactive slider since 2.1.111) or /effort auto to reset. Before escalating between models, consider whether adjusting effort on the current model would suffice:

| Instead of... | Consider... | When |

|--------------|-------------|------|

| Haiku → Sonnet | Stay on Haiku | Task is still deterministic, just needs more context |

| Sonnet → Opus | Opus@medium | Moderate reasoning, not deep architectural analysis |

| Opus@medium → "maybe try again" | Opus@high or "ultrathink" | Genuine complexity needing deeper reasoning |

| Opus 4.7@high → escalate | Opus 4.7@xhigh or @max | Deep architectural analysis on Opus 4.7 specifically |

Default effort change (2.1.68+): Opus 4.6 now

defaults to medium effort for Max and Team

subscribers. Use /model to change effort level, or

type "ultrathink" in your prompt to enable high effort

for the next turn.

Opus 4/4.1 removed (2.1.68+): Opus 4 and 4.1 are

no longer available on the first-party API. Users with

these models pinned are automatically migrated to

Opus 4.6. No action needed for agents using model

frontmatter, as the migration is transparent.

Sonnet 4.5 → 4.6 migration (2.1.69+): Sonnet 4.5

users on Pro/Max/Team Premium are automatically migrated

to Sonnet 4.6. Agent model frontmatter referencing

Sonnet resolves transparently. The --model flags for

claude-opus-4-0 and claude-opus-4-1 now correctly

resolve to Opus 4.6 instead of deprecated versions.

Effort parameter fix (2.1.70+): Fixed API 400 error

This model does not support the effort parameter when

using custom Bedrock inference profiles or non-standard

Claude model identifiers. Effort controls now work

reliably across all deployment configurations.

Default Opus 4.6 on providers (2.1.73+): Bedrock,

Vertex, and Microsoft Foundry now default to Opus 4.6

(was Opus 4.1). Subagent model: opus/sonnet/haiku

aliases now resolve to the current version on all

providers; previously they were silently downgraded to

older versions (e.g., Opus 4.1 instead of 4.6). This

fix means agent dispatch workflows on third-party

providers now match first-party API behavior.

modelOverrides setting (2.1.73+): Maps model

picker entries to provider-specific IDs (Bedrock

inference profile ARNs, Vertex version names, Foundry

deployment names). Use when routing model selections to

specific inference profiles. See the model optimization

guide for configuration details.

/output-style deprecated (2.1.73+): Use /config

instead. Output style is now fixed at session start for

better prompt caching.

Full model IDs in agent frontmatter (2.1.74+): Agent

model: fields now accept full model IDs (e.g.,

claude-opus-4-6) in addition to aliases (opus,

sonnet, haiku). Previously, full IDs were silently

ignored. Agents now accept the same values as --model.

Effort controls do NOT replace the escalation governance

framework: they provide an additional axis. The Iron Law

still applies: investigate before changing either model

or effort level.

Exit Criteria

• [ ] A decision (escalate / stay) is stated with a named trigger from the "When to Escalate" or

"When NOT to Escalate" tables, not a vague claim of complexity.

• [ ] If escalation is recommended, the specific target model and the subtask scope are documented

before the escalation occurs.

• [ ] Any instance of escalation triggered by "maybe a smarter model will figure it out" is flagged

as an Iron Law violation and blocked.

• [ ] If the task is decomposable into smaller pieces that each fit the current model, that

decomposition is proposed instead of escalation.