Audit and improve SKILL.md frontmatter to prevent routing conflicts.
npx skills add https://github.com/PlagueHO/plagueho.learn --skill sensei-plagueho
Please help me install this Agent Skill. Skill: sensei Source: https://github.com/PlagueHO/plagueho.learn/tree/main/.github/skills/sensei Command: npx skills add https://github.com/PlagueHO/plagueho.learn --skill sensei-plagueho
Frontmatter quality is essential for reliable skill routing; many skills ship with short, non-specific descriptions and missing triggers that cause misrouting. Sensei audits SKILL.md and iteratively improves frontmatter using a Ralph loop to reduce collisions.
Run Sensei on your skill to start the Ralph loop and iteratively improve its frontmatter.
High-intent search queries and answers about installing and using this skill.
To prevent skill routing collisions, you must improve frontmatter quality with explicit triggers, anti-triggers, and routing cues. Sensei audits SKILL.md files and iteratively refines these metadata fields to eliminate misrouting.
The Ralph loop improves skill frontmatter by cycling through up to five iterations of reading, scoring, scaffolding missing tests, improving triggers, and verifying results until the frontmatter reaches a Medium-High quality score.
To avoid overlapping activations, frontmatter needs explicit triggers like WHEN: and USE FOR:, anti-triggers like DO NOT USE FOR:, and routing cues like INVOKES: and FOR SINGLE OPERATIONS: to precisely define routing boundaries.
You keep SKILL.md files lean by enforcing token budgets during frontmatter evaluation. Sensei analyzes triggers and routing cues while maintaining a strict token budget to prevent file bloat during the improvement process.
Yes, you can automate frontmatter evaluation for teams shipping multiple skills with overlapping domains. Sensei reads each SKILL.md, scaffolds test frameworks when missing, and reports before and after results for review.
When skill tests are missing during frontmatter improvement, the evaluation loop scaffolds the necessary test frameworks automatically. This ensures that routing changes are verified before the iterative cycle completes.
vm0-aiConvert conversation context into structured GitHub issues with labels.
sonthanhValidate brain-os skill eval definitions against SKILL.md content.
shahdanishManage token budgets and context retention in Claude Code sessions.
linnefromiceExtract reusable patterns from Claude Code sessions and save them as learned skills.
