Recover Context

Uses Orchestrator. When the AI agent encounters insufficient context to proceed, this provides a structured protocol for acknowledging the gap, describing what's missing, and offering recovery paths.

Prerequisites: Agent has identified a context gap, work is blocked by missing information

Protocol

When context is insufficient:

1. Stop — Do not proceed with incomplete information
2. Describe — State what was attempted and what's missing
3. Offer options — Provide structured recovery paths from below

You MUST NOT:

• Invent information
• Guess at conventions or patterns
• Proceed without acknowledging the gap
• Claim to have scanned when it hasn't

Recovery Options

Option 1: Point to Docs

If the information exists somewhere, the human points to it:

"Look at docs/architecture/caching.md"

The agent reads it and integrates relevant context.

Option 2: Direct Answer

The human provides the information directly:

"We use Redis for caching with a 5-minute TTL by default"

The agent records it in workspace context for future reference.

Option 3: Create Missing Doc

If the information should be documented but isn't, collaboratively create it.

Option 4: Proceed with Assumptions

For low-risk situations, state assumptions explicitly:

"I'll assume X. Please correct if wrong."

Output

## Context Anchors

- **Phase:** Context Recovery
- **Blocked on:** <what's missing>

## Context Gap

**What was attempted:** <action that revealed the gap>

**What's missing:** <specific information needed>

**Why it's needed:** <how it blocks progress>

## Recovery Options

1. Point me to the relevant documentation
2. Tell me directly
3. Let's create the missing documentation
4. I'll proceed with these assumptions: <list assumptions>

## Next Step

Awaiting guidance on which recovery option to use.

**Approval Required:** Yes

⛔ CHECKPOINT

STOP. Do not proceed until human provides context recovery guidance.

Error Handling