profit 77655c298c Initial commit: Agent Governance System Phase 8

Phase 8 Production Hardening with complete governance infrastructure:

- Vault integration with tiered policies (T0-T4)
- DragonflyDB state management
- SQLite audit ledger
- Pipeline DSL and templates
- Promotion/revocation engine
- Checkpoint system for session persistence
- Health manager and circuit breaker for fault tolerance
- GitHub/Slack integrations
- Architectural test pipeline with bug watcher, suggestion engine, council review
- Multi-agent chaos testing framework

Test Results:
- Governance tests: 68/68 passing
- E2E workflow: 16/16 passing
- Phase 2 Vault: 14/14 passing
- Integration tests: 27/27 passing

Coverage: 57.6% average across 12 phases

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

2026-01-23 22:07:06 -05:00

2.9 KiB

Raw Blame History

Context Checkpoint Skill

Overview

The checkpoint skill preserves session context across token resets and orchestrates sub-agent calls with minimal token usage.

Commands

Create Checkpoint

/opt/agent-governance/bin/checkpoint now [--notes "description"]

Captures current state: phase, tasks, dependencies, variables, recent outputs.

Load Checkpoint

/opt/agent-governance/bin/checkpoint load [checkpoint_id]

Loads latest or specific checkpoint. Use --json for machine-readable output.

Compare Checkpoints

/opt/agent-governance/bin/checkpoint diff [--from ID] [--to ID]

Shows changes between checkpoints (phase, tasks, variables, dependencies).

List Checkpoints

/opt/agent-governance/bin/checkpoint list [--limit N]

Lists available checkpoints with timestamps and token counts.

Context Summary

/opt/agent-governance/bin/checkpoint summary --level minimal|compact|standard|full

Generates token-aware summaries:

minimal (~500 tokens): Phase + active task + agent
compact (~1000 tokens): + pending tasks + key variables
standard (~2000 tokens): + all tasks + dependencies
full (~4000 tokens): Complete context for complex operations

Auto-Orchestrate Mode

/opt/agent-governance/bin/checkpoint auto-orchestrate --model minimax|gemini|gemini-pro \
    --instruction "command1" --instruction "command2" [--dry-run] [--confirm]

Delegates commands to OpenRouter models with automatic checkpointing.

Instruction Queue

/opt/agent-governance/bin/checkpoint queue list|add|clear|pop [--instruction "..."] [--priority N]

Manages pending instructions for automated execution.

Prune Old Checkpoints

/opt/agent-governance/bin/checkpoint prune [--keep N]

Removes old checkpoints, keeping the most recent N (default: 50).

When to Use

Before complex operations: Create checkpoint to preserve state
After session reset: Load checkpoint to restore context
Sub-agent calls: Use summary --level compact to minimize tokens
Debugging: Use diff to see what changed between checkpoints
Automated mode: Use auto-orchestrate for humanless execution

Examples

# Save current state before risky operation
checkpoint now --notes "Before database migration"

# Restore after context reset
checkpoint load

# Get minimal context for sub-agent
checkpoint summary --level minimal

# Run automated commands with Minimax
checkpoint auto-orchestrate --model minimax \
    --instruction "run tests" \
    --instruction "deploy if tests pass"

# Queue instructions for later
checkpoint queue add --instruction "backup database" --priority 10
checkpoint queue list

Integration

Checkpoints are stored in:

/opt/agent-governance/checkpoint/storage/ (JSON files)
DragonflyDB checkpoint:latest key (fast access)

Audit logs go to:

SQLite: orchestration_log table
DragonflyDB: orchestration:log list

2.9 KiB Raw Blame History