Skip to content
Back to skills

Graph Execution

ASecurity

Executes complex workflows using deterministic graph-state machines, explicit transition guards, transactional worktrees, and receipt gates.

  • 7 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added September 2, 2026
ai-agentspythonbashnodegit

Security analysis

A100/100

Pro scans all 4 files and shows the line behind each finding

Scanned October 3, 2026

npx -y skills add richfrem/agent-plugins-skills --skill graph-execution --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Graph Execution?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Graph Execution
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/richfrem-graph-execution/badge)](https://www.skillsdirectory.com/skills/richfrem-graph-execution)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: graph-execution
plugin: agent-orchestration
description: Executes complex workflows using deterministic graph-state machines, explicit transition guards, transactional worktrees, and receipt gates.
allowed-tools: Bash, Read, Write
---

# Graph-Planned Execution (`graph-execution`)

Deterministic execution primitive for tasks requiring formal state tracking, human approval gates, worktree sandboxing, and safe rollback.

## Contents

- [Constraints](#constraints)
- [Quick start](#quick-start)
- [State Machine](#state-machine)
- [Workflow](#workflow)
- [Verification](#verification)
- [References](#references)

## Constraints

- **State persistence**: Active node and attempt counts persist in `.agent/learning/evolution_state.json`.
- **Proposal mode invariant**: Zero mutations or worktrees are created before explicit human authorization.
- **Verifier sovereignty**: Pre-execution hashes of verifiers are locked; mutation target cannot alter the verifier.
- **Asymmetric persistence**: On 3rd failure, code changes rollback while learnings and debt persist to Layer 2.

## Quick start

```bash
# Query active graph execution state or initiate node transition
python3 plugins/agent-agentic-os/scripts/evolution_state.py status
```

## State Machine

```mermaid
stateDiagram-v2
    [*] --> TRIAGE
    TRIAGE --> PLAN: 4-Box Gate Passed
    PLAN --> AWAITING_APPROVAL: Manifest Formulated
    AWAITING_APPROVAL --> AUTHORIZED: User Explicit Approval
    AUTHORIZED --> CREATE_WORKTREE: Sandbox Initialized
    CREATE_WORKTREE --> EXECUTE: Mutation Attempt (1-3)
    EXECUTE --> VERIFY_GATE: Objective Verifier Run
    VERIFY_GATE --> COMMIT: Verifier Passed (exit 0)
    VERIFY_GATE --> ROLLBACK: Verifier Failed (Attempts == 3)
```

## Workflow

1. **TRIAGE**: Evaluate 4-Box Automation Gate (structural issue, objective verifier, ceiling <= 3, persistence sink).
2. **PLAN**: Formulate Transaction Manifest declaring `mutation_targets`, `verifier`, and `forbidden_paths`.
3. **AUTHORIZED**: Request explicit user authorization; spawn isolated git worktree upon approval.
4. **EXECUTE**: Apply surgical code edits strictly confined inside the isolated worktree sandbox.
5. **VERIFY_GATE**: Run objective verifier command. Advance on exit 0; retry or rollback on exit != 0.
6. **COMMIT / ROLLBACK**: On pass, stage and commit with receipt; on 3rd failure, rollback code and export learnings.

## Verification

```bash
# Verify state machine integrity and receipt logs
python3 plugins/agent-agentic-os/scripts/verify_evolution_receipt.py --audit
# Ensure active worktree clean status
git status --short
```

## References

- [PATTERN_GUIDE.md](references/PATTERN_GUIDE.md) — Comprehensive comparative pattern guide.
- [acceptance-criteria.md](references/acceptance-criteria.md) — Verification criteria and contracts.
- [fallback-tree.md](references/fallback-tree.md) — Recovery and failure escalation paths.

Files in this skill

  • SKILL.md5.1 KB
  • evals/evals.json724 B
  • references/acceptance-criteria.md42 B
  • references/fallback-tree.md36 B

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.