continuous-learning-v2

基于本能的学习系统，通过钩子观察会话，创建具有置信度评分的原子本能，并将其进化为技能/命令/代理。

Quality

50%

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 ./docs/zh-CN/skills/continuous-learning-v2/SKILL.md

Quality

Content

42%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 reads more like a product announcement than an operational skill: setup steps are executable, but the core functionality (observer agent, /instinct commands, evolution clustering) is described in diagrams and tables rather than implemented or referenced, and no bundle files back the referenced scripts. Trimming the promotional sections and shipping the actual hook/CLI code (or references to it) would address all four dimensions.

Suggestions

Remove or compress the v1/v2 comparison table, the "why hooks" philosophy section, backward-compat narrative, and tagline; keep only operational content to recover token budget.

Ship the referenced implementation — hooks/observe.sh and the Python CLI — in a scripts/ directory, or replace the /instinct-* command table with actual command definitions the user can install.

Add a validation checkpoint after hook setup (e.g., run a tool call and confirm observations.jsonl gains a line) so the workflow has a feedback loop before relying on observation data.

Move the full config.json schema and confidence-scoring rules into a references/ file and link to it one level deep from a leaner SKILL.md.

Dimension	Reasoning	Score
Conciseness	The body carries noticeable padding: a v1-vs-v2 marketing comparison table, a "为什么用钩子而非技能进行观察" philosophy section with a pull quote, backward-compatibility narrative, and a closing tagline — content that explains/promotes rather than instructs and could be cut without losing actionable guidance.	2 / 5
Actionability	The hook JSON configs, mkdir commands, and config.json example are concrete and copy-paste ready, but the four /instinct-* commands and the "Python CLI" are named without any implementation — the referenced hooks/observe.sh and CLI do not exist in the bundle (no references/, scripts/, or assets/ directories are present).	3 / 5
Workflow Clarity	The quick start provides a visible three-step sequence (enable hooks → init directories → use commands), but there are no validation checkpoints (e.g., verify hooks fire or that observations.jsonl is growing), and the later stages depend on commands and a CLI whose mechanics are never shown.	3 / 5
Progressive Disclosure	A single ~295-line file inlines material that belongs in separate references — the full config schema, confidence-scoring philosophy, and v1/v2 history — with no pointers to external files, though section headers do provide basic structure.	3 / 5
	Total	11 / 20 Passed

Description

58%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 communicates a clear and fairly specific set of capabilities in third person, but it omits any 'when to use' trigger guidance and relies on internal jargon (instincts, confidence scores, hooks) rather than natural user phrasing. Adding a 'Use when...' clause with plain-language triggers would raise both completeness and trigger-term quality.

Suggestions

Append an explicit trigger clause, e.g. "Use when setting up automatic learning from Claude Code sessions, tuning instinct confidence thresholds, or exporting/importing learned behaviors."

Add natural synonyms users would actually say — e.g. "自动学习" (auto-learning), "会话观察" (session observation), "行为模式" (behavior patterns) — alongside the instinct/hooks jargon.

Clarify scope boundaries (e.g., that it learns coding preferences and workflows from session activity) to reduce overlap with memory and settings-management skills.

Dimension	Reasoning	Score
Specificity	The description lists three concrete actions — "通过钩子观察会话" (observes sessions via hooks), "创建具有置信度评分的原子本能" (creates atomic instincts with confidence scoring), "并将其进化为技能/命令/代理" (evolves them into skills/commands/agents) — with only minor coverage gaps.	4 / 5
Completeness	The "what" is clearly stated (observe sessions, create confidence-scored instincts, evolve into skills/commands/agents), but there is no "when" — no 'Use when...' clause or equivalent trigger guidance, which the guidelines cap at 3.	3 / 5
Trigger Term Quality	Keywords like "钩子" (hooks), "本能" (instincts), and "置信度" (confidence) are domain jargon rather than natural phrases a user would say when needing this skill (e.g., "learn from my sessions", "auto-learn my preferences"), and no synonyms or variations are offered.	3 / 5
Distinctiveness Conflict Risk	The instinct-based learning niche with hooks and confidence scoring is mostly distinct from typical skills; minor overlap risk remains with memory or settings-management skills because the missing trigger guidance leaves its boundaries vague.	4 / 5
	Total	14 / 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: xu-xiang/everything-claude-code-zh
Path: docs/zh-CN/skills/continuous-learning-v2/SKILL.md
Commit: dfbf946

Reviewed: 2 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.