功能描述

当执行完变更需要验证是否应保留、候选被标记 pending 需要人工审批、或想查看待审队列时使用。7 层机械门禁: Schema→Compile→Lint→Regression→Review→Doubt→HumanReview，任一 required 层失败即拒绝。不用于打分（用 improvement-disc...

使用说明 (SKILL.md)

Improvement Gate

Name: Improvement Gate
Author: lanyasheng

7-layer mechanical quality gate: any required layer fail = reject/revert.

When to Use

验证已执行的候选是否应保留（gate.py）
管理人工审核队列（review.py --list）
完成待审批项（review.py --complete）

When NOT to Use

给候选打分 → use improvement-discriminator
执行文件变更 → use improvement-executor
评估 skill 结构 → use improvement-learner

7-Layer Gate

Layer	Gate	Required	Pass Condition
0	SchemaGate	Yes	Candidate has id, category, risk_level, execution_plan
1	CompileGate	Yes	Modified .py files pass `py_compile`; non-Python files auto-pass
2	LintGate	No (advisory)	No lines >120 chars, no mixed tabs/spaces in diff
3	RegressionGate	Yes	Evaluator verdict != "reject" (checks `evaluator_evidence`)
4	ReviewGate	Yes	Discriminator recommendation=accept AND panel not DISPUTED AND LLM judge != reject
5	DoubtGate	Yes	Candidate text has \x3C threshold hedging words (threshold varies by category: docs=2, prompt=4, code=3)
6	HumanReviewGate	No (advisory)	Flags `needs_human=true` for medium/high risk or prompt/workflow/tests/code categories

Gate execution: layers run sequentially. First required-layer failure stops execution and triggers revert (if file was modified) or reject.

CLI — gate.py

python3 scripts/gate.py \
  --ranking ranking.json \         # REQUIRED: ranking artifact from discriminator
  --execution execution.json \     # REQUIRED: execution artifact from executor
  --state-root /path/to/state \    # default: lib/state_machine.DEFAULT_STATE_ROOT
  --evaluation eval.json \         # optional: evaluator artifact (forwarded by orchestrator)
  --layers schema,compile,review \ # optional: run only these layers (default: all 7)
  --output receipt.json            # default: auto-generated path

Param	Default	When to change
`--layers`	all 7	Use `schema,compile` for fast structural checks only
`--evaluation`	None	Orchestrator passes this automatically when evaluator ran

CLI — review.py (Human Review Queue)

# List pending human reviews
python3 scripts/review.py --state-root /path/to/state --list

# Complete a review
python3 scripts/review.py --state-root /path/to/state \
  --complete review-cand-01-docs \
  --decision approve \             # approve | reject
  --reason "低风险文档变更，LGTM" \
  --reviewer engineer-name         # default: cli-user

4-Way Decision Logic (after all required layers pass)

Condition	Decision	Action
`accept_for_execution` + low-risk docs/reference/guardrail + success	keep	File stays modified
`recommendation=reject`	revert	Restore backup, append to veto log
`recommendation=hold` OR non-auto-keep-eligible	pending_promote	Restore backup, create review request
`execution.status=unsupported`	reject	No file change, log reason

如果 HumanReviewGate 标记 needs_human=true，即使 keep eligible 也会升级为 pending_promote。

\x3Cexample> 正确: gate 返回 pending_promote → 查审批队列 → 人工批准 $ python3 scripts/review.py --state-root ./state --list → ID: review-cand-01-docs Category: docs Risk: low Since: 2025-01-15T10:00:00Z $ python3 scripts/review.py --state-root ./state --complete review-cand-01-docs --decision approve --reason "confirmed safe" → Review review-cand-01-docs completed: approve \x3C/example>

\x3Canti-example> 错误: gate 返回 revert 后仍然保留文件变更 → revert 时 gate 自动调用 restore_backup()，原文件已恢复。不要手动跳过。 \x3C/anti-example>

Output — Gate Receipt

{"decision": "keep", "reason": "low-risk docs candidate executed successfully",
 "gate_layers": {"all_passed": true, "layers_run": 7, "layer_results": [...]},
 "rollback": {"attempted": false}, "next_step": "propose_candidates", "next_owner": "proposer"}

Related Skills

improvement-discriminator: ReviewGate checks its panel consensus + LLM verdict
improvement-executor: Gate validates executor output; reverts via rollback_pointer
improvement-evaluator: RegressionGate checks evaluator verdict when --evaluation provided
improvement-orchestrator: Calls gate as stage 5, forwards evaluator artifact
benchmark-store: Pareto front data consumed by RegressionGate

安全使用建议

This skill appears coherent for enforcing a multi-layer quality gate. Before installing or running it: 1) Verify the implementation of the repo-local helpers (lib.state_machine, lib.common) since gate.py calls restore_backup(), update_state(), and writes review/receipt JSON files — those functions control what gets changed on disk. 2) When invoking the gate, point --state-root at a controlled directory and ensure execution/ranking/evaluation artifacts come from trusted orchestrator components; the gate trusts those artifacts (including a rollback_pointer.target_path) and may operate on the target path. 3) Review restore_backup() / rollback behavior to confirm it cannot be abused to overwrite sensitive files if given a crafted artifact. 4) Run the included tests locally to confirm behavior in your environment. Overall the skill does not request unrelated credentials or network installs and is consistent with its stated purpose.

功能分析

Type: OpenClaw Skill Name: auto-improvement-gate Version: 1.0.0 The skill bundle implements a multi-layered quality gate system designed to validate automated improvements to code and documentation. The core logic in `scripts/gate.py` includes structural checks, syntax validation via `py_compile`, and a 'DoubtGate' that scans for non-committal language. The `scripts/review.py` file provides a standard CLI for managing human approval workflows. All behaviors, including file restoration and state management, are consistent with the stated purpose of a quality control mechanism and do not exhibit signs of data exfiltration, malicious execution, or harmful prompt injection.

能力评估

✓ Purpose & Capability

Name/description describe a multi-layer quality gate and the provided scripts (gate.py, review.py) implement schema/compile/lint/regression/review/doubt/human-review layers and a human-review CLI; required inputs (ranking, execution, evaluation artifacts and a state directory) match the stated purpose.

✓ Instruction Scope

SKILL.md instructs the agent to run local Python scripts with artifacts and a state root. The runtime steps reference only local files, review queue state, and orchestrator-provided artifacts. There are no instructions to read unrelated system secrets or send data to third-party endpoints.

✓ Install Mechanism

No install spec — instruction-only with bundled scripts and tests. No downloads, package installs, or external executables are required.

✓ Credentials

The skill declares no required environment variables, credentials, or external config paths. The code operates on artifact JSON and a user-provided state directory, which is proportionate to a gate/review workflow.

✓ Persistence & Privilege

always is false and the skill does not demand permanent platform-level privileges. It reads/writes review files under the provided state-root and may call restore_backup/update_state — file I/O is expected for a gate but means callers should choose a safe state-root.

版本历史

v1.0.0

Initial release: closed-loop skill improvement pipeline

元数据

Slug auto-improvement-gate

版本 1.0.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 1

常见问题