decision-gates

在 DSH 长程任务中为决策层加装五道防偏闸门：原始证据锚、对抗式对齐审计、跨包一致性校验、成本比对、防御自检，防止目标漂移、验收失真与摘要偏差。用于跨会话长任务、多 worker 并行、审计类任务、成本敏感任务的 checkpoint 记录与恢复，以及任何需要"决策层不跑偏"保障的执行场景。

Quality

80%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

81%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The content is a tightly written, well-sequenced multi-gate workflow with explicit validation and feedback loops, scoring high on workflow clarity. Its gaps are moderate: two referenced template files are absent from the bundle, and some trigger-list content is duplicated across sections.

Suggestions

Add the missing bundle files templates/raw_evidence_anchors.json and templates/adversarial-audit-brief.md so the workflow's stage-1/3 and validation steps are executable as written.

De-duplicate the adversarial-audit trigger list: define it once in 核心规则 (闸门 1) and reference it from 阶段 3 instead of restating it verbatim.

Dimension	Reasoning	Score
Conciseness	The body is information-dense and avoids explaining concepts Claude already knows, but the adversarial-audit trigger event list is repeated verbatim in 核心规则 (闸门 1) and 阶段 3, which could be consolidated.	4 / 5
Actionability	Provides concrete commands (python scripts/run_tests.py, gate_scope_lock.py, gate_deadband.py, job_output(..., timeout_ms=60000)), specific thresholds (≥2, ≥5, 连续 3 次), and named JSON fields; however the two referenced template files (templates/raw_evidence_anchors.json, templates/adversarial-audit-brief.md) are absent from the bundle, leaving a moderate executability gap.	4 / 5
Workflow Clarity	Six clearly sequenced phases with explicit validation checkpoints (❌ 冲突不得进入验收, 审计结论 ⚠️/❌ 必须先 resolve, bias ≥2 回退), feedback loops (回退/恢复), and a verification budget checklist — fully satisfying the destructive/batch validation requirement.	5 / 5
Progressive Disclosure	Well-organized into clear sections with one-level-deep, clearly signaled references to scripts/ (all six scripts present), but two referenced templates/ files are missing from the actual bundle, a minor organization/navigation defect.	4 / 5
	Total	17 / 20 Passed

Description

80%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description clearly and completely conveys both capability and trigger conditions for a distinct niche, with concrete named gates. Its main weakness is jargon-heavy trigger language and a gate count (five) that undercounts the seven gates actually defined in the body.

Suggestions

Replace internal jargon (DSH、决策层防偏闸门) with phrases users would naturally say, and add common synonyms to broaden trigger matching.

Reconcile the gate count: either say '七道' to match the body's gates 0–6 or update the description to name scope-lock and deadband alongside the original five.

Dimension	Reasoning	Score
Specificity	Names the domain and lists five concrete gate actions (原始证据锚、对抗式对齐审计、跨包一致性校验、成本比对、防御自检), but states '五道' (five) while the body defines seven gates (0–6), omitting scope-lock and deadband — a minor coverage gap rather than fully comprehensive.	4 / 5
Completeness	Explicitly answers both what (加装五道防偏闸门…防止目标漂移、验收失真与摘要偏差) and when (用于跨会话长任务、多 worker 并行、审计类任务、成本敏感任务的 checkpoint 记录与恢复…) with concrete trigger scenarios via the '用于…' clause.	5 / 5
Trigger Term Quality	Contains relevant terms (checkpoint、多 worker 并行、审计类任务、成本敏感任务), but the language is dominated by internal jargon (DSH、决策层防偏闸门、原始证据锚) that a user would not naturally say, and common variations/synonyms are missing.	3 / 5
Distinctiveness Conflict Risk	Occupies a clear niche (anti-drift decision gates for long DSH tasks) with distinctive triggers (long-running checkpoints, multi-worker acceptance, audit aggregation, cost-sensitive dispatch) and minimal overlap risk with adjacent skills.	5 / 5
	Total	17 / 20 Passed

Validation

93%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: ooooooooooooooooooop/agent-tools
Path: skills/decision-gates/SKILL.md
Commit: 3677cfc

Reviewed: ~8 hours ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.