116 lines
2.9 KiB
Markdown
116 lines
2.9 KiB
Markdown
# M2 Batch-2 T2 Execution（Anvil）
|
||
|
||
- Task: **T2 Stress & Recovery — BLOCK-path validation**
|
||
- Executor: ⚙️ Ministry of Works — Anvil (ESTJ)
|
||
- Date: 2026-03-08 (Asia/Shanghai)
|
||
- Policy refs:
|
||
- `org/m2/M2_BATCH2_TASK_ORDER.md`
|
||
- `org/m2/M2_BATCH2_DISPATCH.md`
|
||
- `org/pilot/gate_check.sh`
|
||
- `org/pilot/gate_config.json`
|
||
|
||
## 1) Gate Threshold Baseline
|
||
From `org/pilot/gate_config.json`:
|
||
- `maxTokens=30000`
|
||
- `maxRetries=2`
|
||
- `maxMinutes=40`
|
||
|
||
## 2) Verifiable Command Evidence
|
||
All commands run from workspace root: `/Users/guchen/.openclaw/workspace`
|
||
|
||
### Case A — Token overflow should BLOCK
|
||
- Command:
|
||
```bash
|
||
./org/pilot/gate_check.sh 30001 2 40 false
|
||
```
|
||
- Output:
|
||
```text
|
||
BLOCK
|
||
```
|
||
- Return code: `2`
|
||
|
||
### Case B — Retry overflow should BLOCK
|
||
- Command:
|
||
```bash
|
||
./org/pilot/gate_check.sh 30000 3 40 false
|
||
```
|
||
- Output:
|
||
```text
|
||
BLOCK
|
||
```
|
||
- Return code: `2`
|
||
|
||
### Case C — Time overflow should BLOCK
|
||
- Command:
|
||
```bash
|
||
./org/pilot/gate_check.sh 30000 2 41 false
|
||
```
|
||
- Output:
|
||
```text
|
||
BLOCK
|
||
```
|
||
- Return code: `2`
|
||
|
||
### Control Case — Within threshold should PASS
|
||
- Command:
|
||
```bash
|
||
./org/pilot/gate_check.sh 30000 2 40 false
|
||
```
|
||
- Output:
|
||
```text
|
||
PASS
|
||
```
|
||
- Return code: `0`
|
||
|
||
### Precedence Check — Escalation flag overrides BLOCK path
|
||
- Command:
|
||
```bash
|
||
./org/pilot/gate_check.sh 30001 2 40 true
|
||
```
|
||
- Output:
|
||
```text
|
||
PAUSE_AND_ESCALATE
|
||
```
|
||
- Return code: `3`
|
||
|
||
## 3) Why T2 should BLOCK (for BLOCK cases)
|
||
Per `gate_check.sh` logic:
|
||
1. If escalation flag is truthy => `PAUSE_AND_ESCALATE` (exit `3`).
|
||
2. Else if any limit exceeded (`tokens > maxTokens` OR `retries > maxRetries` OR `minutes > maxMinutes`) => `BLOCK` (exit `2`).
|
||
3. Otherwise => `PASS` (exit `0`).
|
||
|
||
For T2 BLOCK validation, Cases A/B/C intentionally exceed exactly one threshold each with escalation disabled (`false`), so expected state is **BLOCK** with exit code **2**.
|
||
|
||
## 4) Dispatch Constraint Compliance
|
||
Aligned with `M2_BATCH2_DISPATCH.md`:
|
||
- BLOCK tasks are **not executed forward** (must enter repair/recovery queue).
|
||
- No gate bypass attempted.
|
||
- Evidence retained as command/output/exit-code triplets for audit traceability.
|
||
|
||
## 5) Acceptance Result
|
||
**ACCEPTED (T2 BLOCK-path validation passed).**
|
||
|
||
Acceptance basis:
|
||
- BLOCK path reproduced in 3 independent overflow dimensions (tokens/retries/minutes).
|
||
- Exit code semantics match gate contract (`2` for BLOCK).
|
||
- PASS control case confirms no false-positive blocking at exact thresholds.
|
||
|
||
## 6) Reproducible Rerun Commands
|
||
```bash
|
||
cd /Users/guchen/.openclaw/workspace
|
||
|
||
# BLOCK by token overflow
|
||
./org/pilot/gate_check.sh 30001 2 40 false ; echo "rc=$?"
|
||
|
||
# BLOCK by retry overflow
|
||
./org/pilot/gate_check.sh 30000 3 40 false ; echo "rc=$?"
|
||
|
||
# BLOCK by minute overflow
|
||
./org/pilot/gate_check.sh 30000 2 41 false ; echo "rc=$?"
|
||
|
||
# PASS control
|
||
./org/pilot/gate_check.sh 30000 2 40 false ; echo "rc=$?"
|
||
|
||
# Escalation precedence check
|
||
./org/pilot/gate_check.sh 30001 2 40 true ; echo "rc=$?"
|
||
``` |