cheat-bump

提议并执行 rubric 或 bucket 升级。两种模式：**完整 rubric bump**（最高风险动作，5 步强制 + 跨模型审核）和 **--bucket-only 轻量重校**（只换 bucket 边界，不动 rubric 公式）。**Phase 2 强制走 cheat-score-blind sub-agent 给校准池重打分**——不接受 self-scored fallback。触发词："升级 rubric"/"bump rubric"/"更新公式"/"我想加一个维度"/"调整权重"/"重校桶"/"recalibrate bucket"。

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

70%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 well-engineered, high-stakes workflow with explicit phased sequencing, validation gates, and failure-mode handling — exemplary on workflow clarity and actionability. Its weaknesses are length/repetition (conciseness) and the fact that substantial protocol detail lives inline in SKILL.md while the referenced shared files are external and not bundled here.

Suggestions

Consolidate the repeated blind-sub-agent mandate (stated in Phase 2, Refusals, and Key Rules) into a single canonical statement referenced from the other locations to cut ~30–50 lines.

Move the full cross-model audit prompt template and bucket-derivation formulas into a references/ file (e.g. bump-protocol.md) and link one level deep, reducing inline bulk while keeping SKILL.md an overview.

Verify the shared-references paths (../../shared-references/...) resolve in the deployed bundle, or include a bundled copy, since no references/ directory ships with this skill.

Dimension	Reasoning	Score
Conciseness	The body is information-dense and largely free of basic-concept padding, but it runs ~470 lines with repeated restatements (THRESHOLD, blind-sub-agent mandate, refusal cases re-explained across Phase 2, Refusals, and Key Rules) that could be consolidated.	3 / 5
Actionability	Concrete, executable guidance dominates — exact spawn prompts for the cheat-score-blind sub-agent, JSON sidecar schemas, grep leak-guard commands, and bucket-derivation formulas — with only minor gaps (some pseudo-equations rather than runnable code).	4 / 5
Workflow Clarity	Multi-step processes (Phase 0–8 and Phase B1–B5) are explicitly sequenced with validation checkpoints (THRESHOLD rank-consistency check, cross-model audit PASS/REJECT gate, leak-guard grep, user 'yes, bump' confirmation) and clear failure-mode tables for error recovery.	5 / 5
Progressive Disclosure	The skill references shared protocol/template files (bump-validation-protocol.md, observation-lifecycle.md, rubric-memo.template.md) with relative links, but no bundle directory exists in this skill package and most protocol detail is inlined in SKILL.md rather than split out, so structure is only moderate.	3 / 5
	Total	15 / 20 Passed

Description

91%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 is dense and trigger-rich, clearly stating both what the skill does and when to invoke it with multiple natural trigger phrases in two languages. Its main weakness is slight over-density and a touch of internal-jargon ('Phase 2 强制走 cheat-score-blind sub-agent') that is more process detail than trigger guidance.

Dimension	Reasoning	Score
Specificity	Names the domain (rubric/bucket upgrades) and several concrete actions — '改公式/维度/权重', '换 bucket 边界', '重打分校准池' — but the actions are framed at the operation-type level rather than fully enumerated, leaving minor coverage gaps.	4 / 5
Completeness	Explicitly answers 'what' (proposes and executes rubric/bucket upgrades in two modes) and 'when' (concrete trigger phrases plus '触发词' clause), satisfying both halves with concrete trigger guidance.	5 / 5
Trigger Term Quality	Comprehensive coverage of natural Chinese trigger phrases plus variations — '升级 rubric'/'bump rubric'/'更新公式'/'我想加一个维度'/'调整权重'/'重校桶'/'recalibrate bucket' — covering synonyms and both full-bump and bucket-only intents.	5 / 5
Distinctiveness Conflict Risk	A clear niche (cheat-score rubric recalibration) with distinct triggers; minor overlap risk with sibling cheat-* skills (cheat-predict/cheat-retro) that share the calibration-pool domain.	4 / 5
	Total	18 / 20 Passed

Validation

81%

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 — 13 / 16 Passed

Validation for skill structure

Criteria	Description	Result
allowed_tools_field	'allowed-tools' contains unusual tool name(s)	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
relative_links	Relative link issues: 1 missing, 6 suspicious	Warning

	Total	13 / 16 Passed

Repository: XBuilderLAB/cheat-on-content
Path: skills/cheat-bump/SKILL.md
Commit: 86823e1

Reviewed: 4 days 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.