Executes complex workflows using deterministic graph-state machines, explicit transition guards, transactional worktree isolation, receipt gates, and automatic rollback on verification failure.
Scanned 9/2/2026
Install to Claude Code
npx -y skills add richfrem/agent-plugins-skills --skill graph-execution --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Graph Execution?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/richfrem-graph-execution)More formats (shields.io, HTML) on the badges page.
---
name: graph-execution
plugin: agent-orchestration
description: "Executes complex workflows using deterministic graph-state machines, explicit transition guards, transactional worktree isolation, receipt gates, and automatic rollback on verification failure."
allowed-tools: Bash, Read, Write
---
# Graph-Planned Execution (Deterministic State Machine Protocol)
Provides a deterministic, graph-planned execution primitive for tasks requiring formal state tracking, non-negotiable human approval gates, isolated worktree sandboxing, and guaranteed safe rollback.
---
## 1. Overview: Loops vs. Graphs
While standard agent loops rely on conversational context to decide next steps, **Graph Execution** models tasks as a formal Directed Acyclic Graph (DAG) or finite state machine:
```mermaid
stateDiagram-v2
[*] --> TRIAGE: Task / Error Trigger
TRIAGE --> PLAN: 4-Box Automation Gate Passed
PLAN --> AWAITING_APPROVAL: Transaction Manifest Formulated
AWAITING_APPROVAL --> AUTHORIZED: User Explicit Approval ("Proceed")
AUTHORIZED --> CREATE_WORKTREE: Worktree Sandbox Initialized
CREATE_WORKTREE --> EXECUTE: Mutation Attempt (1-3)
EXECUTE --> VERIFY_GATE: Objective Verifier Run
VERIFY_GATE --> PRE_COMMIT_RECEIPT: Verifier Passed (exit 0)
VERIFY_GATE --> PLAN: Verifier Failed (Attempts < 3)
VERIFY_GATE --> ROLLBACK: Verifier Failed (Attempts == 3)
PRE_COMMIT_RECEIPT --> COMMIT: Staged Tree Validated & Committed
ROLLBACK --> FINAL_RECEIPT: Layer 2 Transferred & Worktree Removed
COMMIT --> FINAL_RECEIPT: Local Cycle Successful
FINAL_RECEIPT --> [*]: Terminal Receipt Logged
```
---
## 2. The Core Invariants
1. **State in Files, Not Prompts:**
The active node, attempt count, and operation bounds are stored in `.agent/learning/evolution_state.json` and mediated by a deterministic controller script (e.g., `evolution_state.py`).
2. **Proposal Mode Invariant (Read-Only Planning):**
Prior to explicit user approval (`AWAITING_APPROVAL`), the agent executes **zero state-changing operations**, creates **no git worktrees**, and runs **no mutations**.
3. **Transactional Worktree Sandboxing:**
Mutations are executed inside an isolated git worktree (`../worktree-evolution-<cid>`), keeping the main checkout 100% clean during execution.
4. **Verifier Sovereignty:**
The agent cannot modify the test scripts, holdout sets, or evaluation criteria that judge its own work. Pre-execution hashes of verifiers are locked.
5. **Asymmetric Persistence:**
If a task fails after maximum retry attempts (e.g., 3 attempts):
- **Code changes are safely rolled back** and the worktree is deleted.
- **Failure learnings, negative constraints, and reproduction notes** are permanently preserved in Layer 2 Markdown playbooks and debt logs before worktree teardown.
6. **Integrity Receipts:**
No commit is permitted without an **Evolution Integrity Receipt** (`EVO-INTEGRITY-<cid>-<hash>`) binding the exact staged git tree to the audit event sequence.
---
## 3. The Canonical Graph Node Lifecycle
### Node 1: `TRIAGE` (Intake & Qualification)
- Ingests the task, error signature, or trigger.
- Evaluates the **4-Box Automation Gate**:
1. Is the problem structural or recurring?
2. Is there an objective programmatic verifier (`exit 0`)?
3. Is there an iteration ceiling (max 3 attempts)?
4. Is there an immutable persistence sink for learnings?
### Node 2: `PLAN` (Proposal Mode)
- Formulates an immutable **Transaction Manifest** declaring:
- `mutation_targets`: Exact relative file paths to edit.
- `verifier`: The objective verification command (`argv` array).
- `forbidden_paths`: Protected files (verifiers, test runners, root policies).
- Halts at `AWAITING_APPROVAL` and presents the plan to the human.
### Node 3: `AUTHORIZED` & `CREATE_WORKTREE`
- Entered **only** upon explicit user confirmation ("Proceed").
- Spawns the isolated worktree (`git worktree add -b evolution/<cid> ../worktree-evolution-<cid>`).
- Captures baseline untracked files.
### Node 4: `EXECUTE` (Surgical Mutation)
- Applies surgical edits against declared `mutation_targets` inside the worktree.
- Records all created or modified files.
### Node 5: `VERIFY_GATE` (Objective Proof Check)
- Validates verifier hashes are unmodified.
- Runs the verifier command in non-mutating mode.
- If `exit 0`: transitions to `PRE_COMMIT_RECEIPT`.
- If `exit != 0` and attempts < 3: increments attempt count and transitions to `PLAN`.
- If `exit != 0` and attempts == 3: transitions to `ROLLBACK`.
### Node 6: `COMMIT` or `ROLLBACK` (Asymmetric Resolution)
- **On Pass:** Persists confirmed knowledge to Layer 2 wiki $\rightarrow$ Stages files $\rightarrow$ Generates Pre-Commit Receipt $\rightarrow$ Commits.
- **On 3rd Failure:** Transfers failure insights to main checkout $\rightarrow$ Removes worktree $\rightarrow$ Logs open debt item.
### Node 7: `FINAL_RECEIPT` (Closure)
- Emits final receipt token and marks cycle `COMPLETED` or `ESCALATED`.
---
## Operational References
See [`references/PATTERN_GUIDE.md`](../references/PATTERN_GUIDE.md) and [`references/acceptance-criteria.md`](references/acceptance-criteria.md) for complete state machine contracts.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!