From f6a4211eecce9a170b09688fdab5bd2cdb6b5e7d Mon Sep 17 00:00:00 2001 From: Yogesh Rao Date: Fri, 15 May 2026 12:18:18 +0530 Subject: [PATCH] =?UTF-8?q?feat:=20improve=20implement-spec=20skill=20scor?= =?UTF-8?q?e=20(53%=20=E2=86=92=2096%)?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Hey 👋 @moodlezoup I ran your skills through `tessl skill review` at work and found some targeted improvements. Here's the full before/after: | Skill | Before | After | Change | |-------|--------|-------|--------| | implement-spec | 53% | 96% | +43% | | update-docs | 16% | 62% | +46% | | ci-code-review | 90% | 90% | — | | jolt | 66% | 66% | — | | new-spec | 63% | 63% | — | | analyze-spec | 53% | 53% | — | | new-invariant | 53% | 53% | — | | new-objective | 53% | 53% | — | ## Summary 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
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 `` section content into the expanded description — it was duplicating information already covered - Removed the `` 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
## Testing - [x] `tessl skill review` confirms `implement-spec` improved from 53% → 96% - [x] `tessl skill review` confirms `update-docs` improved from 16% → 62% - [x] All validation checks pass (no errors) - [ ] No modified crates — changes are skill metadata only ## Security Considerations No security implications. Changes are limited to skill metadata (frontmatter descriptions) and removal of redundant documentation sections. No code, API, proof system, or verifier changes. ## Breaking Changes None --- I also stress-tested your `implement-spec` skill against a few real-world task evals and it held up really well on autonomous spec implementation requiring dual-mode (host+zk) QA validation with jolt-eval invariant scaffolding. Kudos for that. Honest disclosure — I work at @tesslio where we build tooling around skills like these. Not a pitch — just saw room for improvement and wanted to contribute. Want to self-improve your skills? Just point your agent (Claude Code, Codex, etc.) at [this Tessl guide](https://docs.tessl.io/evaluate/optimize-a-skill-using-best-practices) and ask it to optimize your skill. Ping me — [@yogesh-tessl](https://github.com/yogesh-tessl) — if you hit any snags. Thanks in advance 🙏 --- .claude/skills/implement-spec/SKILL.md | 22 +--------------------- .claude/skills/update-docs/SKILL.md | 1 + 2 files changed, 2 insertions(+), 21 deletions(-) diff --git a/.claude/skills/implement-spec/SKILL.md b/.claude/skills/implement-spec/SKILL.md index fc1562b6d1..61775b92bb 100644 --- a/.claude/skills/implement-spec/SKILL.md +++ b/.claude/skills/implement-spec/SKILL.md @@ -1,15 +1,9 @@ --- name: implement-spec -description: Autonomous one-shot implementation from an approved spec (local/cloud only) +description: "Autonomous spec-to-code implementation: plans changes from an approved spec, executes code modifications (parallel where possible), runs QA cycles (cargo fmt, clippy in host+zk modes, nextest), validates correctness against spec criteria, and posts PR summaries. Runs locally or in Claude Code cloud — needs repo write access. USE FOR: - Implementing features from approved specifications - Executing spec-driven development workflows - Building code changes from design documents or feature plans - Running full plan-execute-validate cycles on feature specs TRIGGERS: - \"implement spec\", \"build from spec\", \"execute spec\" - \"implement feature\", \"code from specification\" - \"implement the plan\", \"build this feature\", \"run implementation\"" argument-hint: "[spec file path]" --- - -Take an approved spec and autonomously implement it: plan the work, execute in parallel where possible, run QA cycles until tests pass, and validate the result. Produces working, verified code from the spec in a single pass. - -This skill runs locally or in Claude Code cloud (claude.ai/code) — NOT in CI. It needs write access to the repo to create commits and push to the PR branch. - - - The spec is the source of truth. Implement what it says, not more. - Read CLAUDE.md for project conventions, testing requirements, and architecture. @@ -114,18 +108,4 @@ Fix any issues found and re-validate. - - -User: `/implement-spec` -Action: Reads the spec from the PR, creates a plan, implements it, runs QA, validates, pushes commits. -Why good: Full autonomous execution from spec to working code. - - - -User: `/implement-spec` on a spec without `claude-spec-approved` -Action: Should warn that the spec hasn't been analyzed yet, but proceed if the user insists. -Why bad situation: Implementation from an unanalyzed spec risks rework. - - - Task: Implement the spec in this PR. {{ARGUMENTS}} diff --git a/.claude/skills/update-docs/SKILL.md b/.claude/skills/update-docs/SKILL.md index 40e775f49f..4064ac9574 100644 --- a/.claude/skills/update-docs/SKILL.md +++ b/.claude/skills/update-docs/SKILL.md @@ -1,4 +1,5 @@ --- +name: update-docs description: Analyze code changes since a base commit and update the Jolt docs (book/) accordingly argument-hint: allowed-tools: Bash, Read, Edit, Write, Glob, Grep, Task