continuous-learning

Automatically extract reusable patterns from Claude Code sessions and save them as learned skills for future use.

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-TW/skills/continuous-learning/SKILL.md

Quality

Content

46%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 has a solid operational skeleton — concrete config, a pattern-type table, and a working hook snippet — but roughly a third of it is pasted-in research notes that belong in the spec file it cites, and every file reference it makes (the script and the spec doc) dangles because the bundle ships no other files. Cutting the research section and shipping or removing the referenced files would move both progressive disclosure and conciseness up a level.

Suggestions

Delete the "比較筆記（研究：2025 年 1 月）" section (or move it into an actual docs/continuous-learning-v2-spec.md file that ships with the bundle) — it is research notes, not skill instructions.

Ship scripts/evaluate-session.sh (or remove the hook setup that invokes it) and include the actual extraction logic or a worked example of a learned-skill file so the core mechanism is actionable.

Add validation checkpoints to the 運作方式 workflow: what happens when a session is below min_session_length, and how auto_approve: false results are surfaced for user confirmation.

Dimension	Reasoning	Score
Conciseness	The operational core (config block, pattern-type table, hook snippet) is lean, but the trailing "比較筆記" research section (~28 lines of Homunculus comparison, quotes, and v2 enhancement ideas), the "為什麼用 Stop Hook？" rationale, and a tweet link are notebook padding that earns nothing during skill execution.	3 / 5
Actionability	The config JSON and settings.json hook snippet are copy-pasteable, but the actual mechanism — what evaluate-session.sh does, how patterns are detected and written, what a learned-skill file looks like — is absent, and the referenced script is not shipped in the bundle.	3 / 5
Workflow Clarity	"運作方式" lists a three-step sequence (evaluate session → detect patterns → extract skills) but there are no validation checkpoints, no handling for the auto_approve: false path or empty/insufficient sessions, and the executing script is unverifiable.	3 / 5
Progressive Disclosure	The body references docs/continuous-learning-v2-spec.md and ~/.claude/skills/continuous-learning/evaluate-session.sh, neither of which exists in the bundle, while the v2 comparison content that belongs in that spec file is inlined into SKILL.md behind a trailing "---" separator.	2 / 5
	Total	11 / 20 Passed

Description

53%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 gives a clear, third-person "what" with two concrete actions and a reasonably distinct niche, but it entirely lacks "when to use it" guidance and misses the natural trigger vocabulary users would employ. Adding a "Use when..." clause and synonyms would lift three dimensions at once.

Suggestions

Append an explicit trigger clause, e.g. "Use when a session ends having solved recurring problems, or when the user says 'learn from this' / 'remember this pattern'."

Broaden trigger vocabulary with natural synonyms: "learn from sessions", "remember fixes", "save learnings", "self-improve".

Mention one or two more concrete capabilities (e.g. "runs as a Stop hook", "detects error resolutions, workarounds, and user corrections") to round out specificity.

Dimension	Reasoning	Score
Specificity	"extract reusable patterns from Claude Code sessions" and "save them as learned skills" name the domain plus two concrete actions, but coverage stops there — no mention of the evaluation pipeline, pattern types, or configuration that the skill actually offers.	3 / 5
Completeness	The "what" is clear (extract patterns, save as learned skills), but there is no "Use when..." clause or equivalent trigger guidance, which caps completeness at 3 per the rubric guidelines.	3 / 5
Trigger Term Quality	Keywords like "reusable patterns", "Claude Code sessions", and "learned skills" are relevant, but the natural phrases a user would actually say ("learn from this session", "remember this", "save what you learned") are missing.	3 / 5
Distinctiveness Conflict Risk	"Claude Code sessions" + "learned skills" carves a fairly distinct niche in session post-processing, with only minor overlap risk against memory or CLAUDE.md-writing skills.	4 / 5
	Total	13 / 20 Passed

Validation

100%

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 — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: affaan-m/ECC
Path: docs/zh-TW/skills/continuous-learning/SKILL.md
Commit: c70874f

Reviewed: 1 day 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.