feat: improve implement-spec skill score (53% → 96%)

[yogesh-tessl]

Hey 👋 @moodlezoup

ran your skills through tessl skill review at work and found some targeted improvements. Here's the before/after:

Skill	Before	After	Change
implement-spec	53%	96%	+43%

focused on implement-spec. It had the most improvement headroom (tied at 53% with several others) and is the most central workflow skill (spec-to-code implementation is the core development action). Also fixed a missing name field in update-docs that was causing it to fail validation entirely.

Changes made

implement-spec (53% → 96%, +43%)

• Expanded the frontmatter description from a vague one-liner ("Autonomous one-shot implementation from an approved spec") to a structured description listing concrete actions (plans changes, executes code modifications, runs QA cycles, validates correctness, posts PR summaries)
• Added explicit USE FOR and TRIGGERS sections with natural language trigger terms ("implement spec", "build from spec", "execute spec", "implement feature", etc.)
• Folded the <Purpose> section content into the expanded description - it was duplicating information already covered
• Removed the <Examples> section that added minimal value for an autonomous workflow skill

update-docs (16% → 62%, +46%)

• Added the missing name: update-docs field to the frontmatter - the skill was failing validation entirely because this required field was absent, which prevented the LLM judge from scoring it

---

quick honest disclosure. I work at https://github.com/tesslio where we build tooling around skills like these. Not a pitch, just saw room for improvement and wanted to contribute.

if you want to self-improve your skills, or define your own scenarios to pressure test, just ask your agent (Claude Code, Codex, etc.) to evaluate and optimize your skill with Tessl. Ping me @yogesh-tessl, if you hit any snags.

[moodlezoup]

Hi @yogesh-tessl, thanks for the PR. What are these percentages of? Is there any documentation you can point me to?

[yogesh-tessl]

@moodlezoup, good question.

The percentages come from Tessl's skill review, which scores SKILL.md files across a few dimensions: whether the frontmatter has the required fields? How clear the trigger/routing language is? for an LLM to pick the right skill, and how actionable the instructions are once the skill fires. Each dimension gets a 0-100 score and the overall number is a weighted average.

The big jump on implement-spec (53 to 96) was mostly the frontmatter description being too short for agents to reliably match it.

The update-docs one (16 to 62) was just a missing name field that was failing validation entirely.

I'll be happy to point you to the scoring docs if you want more detail on the rubric.