AgentSkillsCN

helix-query

按语义搜索Helix记忆库，返回相关洞察，并按照相关性×有效性×时效性的综合评分进行排序。

SKILL.md
--- frontmatter
name: helix-query
description: Search helix memory by meaning. Returns relevant insights ranked by relevance × effectiveness × recency.
argument-hint: <search text>

Memory Query

Search the learning system for relevant insights using semantic similarity.

Usage

code
/helix-query authentication patterns
/helix-query "database connection errors"

Execution

bash
HELIX="$(cat .helix/plugin_root)"
python3 "$HELIX/lib/memory/core.py" recall "$ARGUMENTS" --limit 10

Output

Display each insight with:

  • •name: Unique identifier (kebab-case slug)
  • •content: Full insight text
  • •tags: Category tags (e.g., debugging, pattern, eval)
  • •effectiveness: 0-1 score based on feedback history
  • •_relevance: Cosine similarity to query (0-1)
  • •_recency: Time decay score (0-1)
  • •_score: Combined ranking score

Scoring Formula

code
score = (0.5 * relevance) + (0.3 * effectiveness) + (0.2 * recency)

Insights that are semantically similar to the query, have helped in past tasks, and were recently used rank highest.

Filtering

bash
# Only return insights with effectiveness > 0.3
python3 "$HELIX/lib/memory/core.py" recall "$ARGUMENTS" --limit 10 --min-effectiveness 0.3