CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

memory-audit-belief-duel

信念对决。当父子节点内容冲突、或两条你都认可的记忆逻辑上不能并存时使用。

Quality

66%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./docs/skills/memory-audit-belief-duel/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

67%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 body delivers a genuinely actionable dual procedure — a numbered duel workflow with fact-vs-inference verification for parent-child conflicts, and reflective interrogation questions for sibling contradictions — anchored by concrete rules and a named tool. Its main weaknesses are rhetorical padding in the first form and the absence of a worked rewrite example that would make the update_memory step copy-ready.

Suggestions

Trim the motivational framing in Form one (e.g., the '如果用户此刻在看着你' and '不舒服了' passages) to the single most effective prompt, saving roughly a third of that section's tokens.

Add one short worked example of a rewritten parent node showing the expected first-person structure (old judgment basis, the fact that broke it, current understanding).

Give Form one the same numbered-step treatment as Form two so both conflict shapes have an explicit sequence with verification checkpoints.

DimensionReasoningScore

Conciseness

Passages such as "不是扫一眼,是回答——用你自己的话,对着你自己说" and "如果这些问题让你不舒服了——好,那你现在的状态是对的" are rhetorical padding that could be tightened, though it never explains concepts Claude already knows and some of the confrontational tone is functional for a self-deception-audit skill. Mostly effective but could be tightened, rather than level 2's 'several unnecessary explanations'.

3 / 5

Actionability

Concrete decision rules ("先核对观察和出处,再修改失效的解读;若事实记录本身也错了,就连同依据一起修正"), a named tool (`update_memory`), an explicit marker to avoid ("[APPEND]"), and specific branches ("不会就删除。会就合并到一条记忆里") give mostly executable guidance. Minor gaps: no worked example of a rewritten parent node and no pointer to where the outcome is recorded.

4 / 5

Workflow Clarity

Form two provides a clear 5-step numbered sequence with verification checkpoints ("先分清哪部分是已核实的事实", step 2's check of supporting records) and a post-rewrite review of dependent child nodes, satisfying validation for the destructive edit/delete operations. Form one is unnumbered reflective bullets without explicit checkpoints, which keeps this at anchor 4 rather than 5.

4 / 5

Progressive Disclosure

No bundle files exist and none are needed; the ~55-line body is cleanly split into two labeled forms with headers and no nested references. It narrowly misses the under-50-line simple-skill criterion and is dual-purpose (two distinct workflows), so structure is good but not maximal.

4 / 5

Total

15

/

20

Passed

Description

48%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 has a clear, well-conditioned trigger clause but never states what the skill does — '信念对决' is a name, not a capability statement. Trigger terms are on-domain but lack synonyms. Adding 2-3 concrete action verbs and natural keyword variations would move it from mediocre to strong.

Suggestions

State the 'what' explicitly with concrete actions, e.g. 'Resolves logically contradictory memory nodes: verifies observations, rewrites outdated parent judgments with update_memory, merges or deletes losing entries.'

Broaden trigger coverage with natural variations users would actually say, such as 'contradictory memories', 'memory conflict', 'outdated belief', or 'inconsistent knowledge'.

Consider an English (or bilingual) gloss in the description so the skill is discoverable when the surrounding conversation is not in Chinese.

DimensionReasoningScore

Specificity

The description names the domain ("信念对决" plus memory-node conflict conditions) but lists zero concrete actions — there are no verbs describing what the skill actually does (audit, rewrite, merge, etc.). It matches the anchor 'Names the domain but actions are minimal or generic' rather than level 3, which requires 1-2 concrete actions.

2 / 5

Completeness

It has an explicit and concrete 'when' ("当父子节点内容冲突、或两条你都认可的记忆逻辑上不能并存时使用") but the 'what' is only a concept label ("信念对决") with no statement of capability. This sits between anchor 2 ('only when present without what') and anchor 3; the concrete double-conditioned trigger pulls it to 3.

3 / 5

Trigger Term Quality

Phrases like "父子节点内容冲突" and "两条你都认可的记忆逻辑上不能并存" are natural trigger terms for this domain, but coverage is thin — no synonyms or variations (e.g., contradictory beliefs, outdated judgment, memory conflict). Some relevant keywords but missing common variations.

3 / 5

Distinctiveness Conflict Risk

The niche is specific — resolving logical contradictions between memory nodes, including parent-child conflicts — and unlikely to trigger unrelated skills. Minor overlap risk remains with sibling memory-maintenance skills (e.g., generic memory cleanup), matching 'Mostly distinct; minor overlap risk with closely related skills'.

4 / 5

Total

12

/

20

Passed

Validation

100%

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

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository

Dataojitori/nocturne_memory

Path

docs/skills/memory-audit-belief-duel/SKILL.md

Reviewed

Table of Contents

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.