CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

auto-memory

Auto-memory guidance — each memory type with examples, the metadata rules and team memory. Load once you decide to save a memory.

Quality

75%

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 ./packages/core/src/skills/bundled/auto-memory/SKILL.md
SKILL.md
Quality
Evals
Security

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.

This is a well-organized, actionable reference that mostly respects the reader's competence: concrete rules, fixed category lists, and paired save-examples per memory type, with verification guidance before recommending from memory. Its main gaps are a missing full memory-file example, a deferred save procedure, and a couple of sections of trimmable meta-explanation.

DimensionReasoningScore

Conciseness

The body is rule-dense with hard limits ("near or below 1,200 characters", "at most 64 characters", "2-6 discriminative retrieval terms") and avoids re-teaching basics, but the intro framing and the "Memory and other forms of persistence" section re-explain plan/task distinctions that could be trimmed — minor over-explanation, matching anchor 4 rather than the lean-every-token 5.

4 / 5

Actionability

Guidance is concrete and executable for an instruction-only skill: a fixed 21-item category list, per-type <examples> pairing user utterances with what to save, an absolute-date conversion example ("Thursday" → "2026-03-05"), and YAML quoting rules. Scored 4 not 5 because no complete worked example of a memory file (frontmatter plus body) is shown, and some rules like "keep keywords unique case-insensitively" lack an illustrative sample.

4 / 5

Workflow Clarity

The reference flows from when-to-save through what-not-to-save, how-to-write, team routing, when-to-access, and verification, with explicit checkpoints ("check the file exists", "grep for it", "update or remove the stale memory"). Not 5 because the actual save procedure ("the two save steps") is deferred to the host instructions, so the end-to-end sequence is not fully present in this document.

4 / 5

Progressive Disclosure

With no bundle files, the ~145-line body is organized into clearly headed, navigable sections; scored against actual bundle structure there are no broken or buried references. Not 5 because it exceeds the under-50-line simple-skill case and the four <type> blocks with examples are bulk content that could plausibly live in a separate reference file.

4 / 5

Total

16

/

20

Passed

Description

62%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 short, third-person, and names its niche clearly, with an explicit load condition that satisfies the 'when' requirement. Its weaknesses are a topic-inventory 'what' rather than concrete actions and a thin set of natural trigger phrases covering only save-time usage.

Suggestions

Rewrite the 'what' as concrete actions, e.g., "Classifies and saves memories across user, feedback, project, and reference types; applies metadata, category, and keyword rules; routes entries between user, project, and team directories."

Broaden trigger terms to cover recall as well as save, e.g., "Load when saving or recalling a memory, when the user says remember/forget, or when maintaining the memory index."

Add the distinction from ordinary persistence ("for durable cross-conversation facts, not in-session state") to sharpen conflict resistance against plan/task-tracking skills.

DimensionReasoningScore

Specificity

"Auto-memory guidance — each memory type with examples, the metadata rules and team memory" names the domain and its content areas, but these are topic labels rather than the several concrete actions (e.g., extracts/fills/merges) that the 4-5 anchors reward.

3 / 5

Completeness

"Load once you decide to save a memory" provides an explicit 'when', and the 'what' (memory types, metadata rules, team memory) is present but framed as a content inventory rather than concrete capabilities; the 'when' also covers only saving, not recall — both present with room to be more specific.

4 / 5

Trigger Term Quality

Terms like "memory", "save a memory", "team memory", and "metadata" are relevant, but common natural variations users would say — "remember", "recall", "forget" — are absent, matching the "missing common variations or synonyms" anchor.

3 / 5

Distinctiveness Conflict Risk

"Auto-memory guidance" occupies a clear niche with a distinct load condition ("once you decide to save a memory"), leaving only minor overlap risk with closely related memory/recall skills; not 5 because the trigger scope is narrow and could still collide with other memory-handling guidance.

4 / 5

Total

14

/

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

QwenLM/qwen-code

Path

packages/core/src/skills/bundled/auto-memory/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.