proactive-memory-judgment

Use when 人物线索需在 F276 即时提案、known-person defer、abstention 间判断。Not for 裸人名、taste、后台扫描或 workspace alias。Output: 可拒绝 proposal、无正文 receipt 或 abstention；永不静默物化。

Quality

65%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./cat-cafe-skills/proactive-memory-judgment/SKILL.md

Quality

Content

75%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 is a well-structured, mostly lean policy skill: an ordered five-gate decision flow, concrete tool signatures, an abstention reasonCode table, and an exit routing section. Its main gaps are minor verbosity from bilingual restatement and absence of explicit validate-retry feedback loops for the canonical-memory write path.

Suggestions

Add an explicit feedback loop for proposal failures (e.g. 'if cat_cafe_propose_person_memory returns an error, re-run the gates before recording an abstention; never treat a failure receipt as a success').

Consolidate the bilingual restatements (defer is not a trash-bin appears in both gate 3 and gate 5) into one canonical statement to trim tokens.

Spell out the full accepted-field set for cat_cafe_propose_person_memory to match the completeness of the abstention and defer signatures.

Dimension	Reasoning	Score
Conciseness	The body is dense and mostly efficient, assuming Claude's competence and avoiding basic-concept padding, but the bilingual prose and a few restated caveats (e.g. defer-not-a-trash-bin repeated across gates 3 and 5) could be trimmed slightly.	4 / 5
Actionability	Concrete tool-call signatures (cat_cafe_record_proactive_memory_abstention, cat_cafe_defer_person_memory_delta with exact accepted/rejected fields) and a reasonCode table give mostly copy-paste-ready guidance, with minor gaps around the propose call's full field set.	4 / 5
Workflow Clarity	The five-gate ordered sequence with an exit routing section and 'fail closed' / immutable-replacement checkpoints is clear and well-sequenced, though explicit validate-then-retry feedback loops for failed proposals are only briefly implied rather than spelled out.	4 / 5
Progressive Disclosure	Well-organized into labeled sections (五道门, Common Mistakes, 出口) with clear hierarchy and no nested external references; as a single self-contained file it is appropriately structured, though at ~100 lines a small amount of the gate rationale could live in a reference file.	4 / 5
	Total	16 / 20 Passed

Description

55%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 explicitly provides both a trigger clause and an output contract, and carves out a distinct niche with anti-triggers. Its weaknesses are heavy reliance on internal jargon (F276, cat_cafe vocabulary, untranslated Chinese terms) instead of natural user language, and second-person phrasing that the rubric penalizes.

Suggestions

Rewrite trigger terms in natural user-facing language (e.g. 'the user mentions a person who matters for future continuity') and keep F276/cat_cafe identifiers as secondary, not the primary trigger.

Convert the description to consistent third-person voice ('Decides whether to ...' rather than 'Not for ...') and merge the bilingual fragments into one clear language.

Add the concrete trigger synonyms users might actually say (e.g. 'remember this person', 'this person will come up again') alongside the abstract continuity-value phrasing.

Dimension	Reasoning	Score
Specificity	Names the domain (person-clue judgment between F276 proposal / defer / abstention) and concrete output forms ('可拒绝 proposal、无正文 receipt 或 abstention'), but the actions are framed as judgment outcomes rather than multiple discrete concrete operations, so coverage is partial rather than comprehensive.	3 / 5
Completeness	Explicitly answers 'when' via the 'Use when ...' clause and 'what' via the 'Output:' clause, but the 'when' leans on jargon and the anti-triggers ('Not for ...') so it could be more explicit and concrete.	4 / 5
Trigger Term Quality	Triggers rely on domain jargon ('人物线索', 'F276 即时提案', 'known-person defer', '裸人名', '后台扫描') rather than natural phrases a user would actually say; '人物线索' is the only user-adjacent term and the rest are internal vocabulary unlikely to be spoken naturally.	2 / 5
Distinctiveness Conflict Risk	The narrow niche (person-clue memory gating between proposal/defer/abstention) plus explicit 'Not for ...' exclusions make it largely distinct from other skills, with only minor overlap risk against general entity/memory skills.	4 / 5
	Total	13 / 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: zts212653/clowder-ai
Path: cat-cafe-skills/proactive-memory-judgment/SKILL.md
Commit: 61389cc

Reviewed: 2 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.