feat: improve skill scores for 5 perseus security skills#3
Open
rohan-tessl wants to merge 1 commit into
Open
Conversation
Hey @kaivyy 👋 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 | |-------|--------|-------|--------| | perseus-audit | 18% | 85% | +67% | | perseus-start | 18% | 84% | +66% | | perseus-scan | 18% | 80% | +62% | | perseus-report | 18% | 76% | +58% | | perseus-exploit | 18% | 74% | +56% | This PR covers **5 of your 15 skills** to keep the contribution focused and reviewable. The remaining 10 skills can be improved in follow-ups or via automated review on future PRs. <details> <summary>Changes summary</summary> **All 5 skills (audit, exploit, report, scan, start):** - Fixed `name` field from colon format (`perseus:audit`) to valid kebab-case (`perseus-audit`) — this was blocking the LLM judge entirely, causing all 5 to score 18% - Expanded frontmatter `description` with specific trigger terms, concrete actions, and explicit "Use when..." clauses - Converted descriptions from bare strings to properly quoted YAML strings - Streamlined introductory paragraphs to remove redundancy with the description field **report skill:** - Extracted the ~300-line inline report template to a new `REPORT_TEMPLATE.md` reference file for better progressive disclosure - Added concrete scoring formula example (`9.5 × 1.0 × 1.0 × 0.9 = 8.55`) - Consolidated scoring factor tables for clarity - Removed inline glossary (standard security terms Claude already knows) - Added validation checkpoint before finalizing **start skill:** - Extracted all detection tables (language, framework, infrastructure, API, AI/LLM) and specialist trigger rules to a new `DETECTION.md` reference file — eliminating duplicate tables that appeared twice in the original - Removed the all-checkmarks language coverage matrix (no information value) - Added error handling guidance ("if any phase fails, log and continue") - Added natural trigger terms (pentest, security assessment, vulnerability scan) to description - Reduced from ~414 lines to ~107 lines while preserving all functional content </details> ## Want your remaining skills optimised too? 🚀 This PR covers **5 of your 15 skills** to keep the contribution focused and reviewable. We have tooling that can go further: - **Optimize remaining 10 skills** automatically (same AI-powered pass as above) - **Add a GitHub Action** — [tesslio/skill-review-and-optimize](https://github.com/tesslio/skill-review-and-optimize) — that automatically reviews (and optionally optimizes) any `SKILL.md` changed in future PRs. Review mode works with zero secrets; maintainers can add `TESSL_API_TOKEN` for AI suggestions and the `/apply-optimize` comment flow. Interested? Just tick the box below and we'll raise a follow-up PR: - [ ] **Yes please!** Raise a follow-up PR: optimize the remaining 10 skills + add the Tessl skill-review-and-optimize GitHub Action - [ ] **No thanks** — happy with the 5 skills in this PR --- 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 - [@rohan-tessl](https://github.com/rohan-tessl) - if you hit any snags. Thanks in advance 🙏
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Hey @thedaviddias 👋
I ran your skills through
tessl skill reviewat work and found some targeted improvements. Here's the full before/after:This PR covers 5 of your 8 skills to keep the contribution focused and reviewable — the remaining 3 (agent-browser at 90%, github-gh at 95%, template-skill placeholder) are already in great shape or intentionally blank.
Changes summary
split-into-references: Added missing YAML frontmatter with name, description, and explicit trigger terms. The skill previously failed validation entirely (no frontmatter = 0% on all judge dimensions).
tdd-methodoly-expert: Removed ~220 lines of verbose TDD textbook content that Claude already knows (Red-Green-Refactor explanations, phase principles, example flows). Consolidated into a tight workflow with explicit validation checkpoints at each phase. Kept the bundled tools section (most valuable part) and added well-signaled references to existing reference files.
brainstorming: Removed the dot graph (won't render in Claude's context), eliminated the triple-redundancy (checklist + process flow diagram + "The Process" section all described the same workflow). Consolidated into a single clear 7-step workflow. Expanded description with specific actions and natural trigger terms.
mcp-builder: Cut ~245 lines of generic software engineering advice (DRY, async/await, error handling concepts Claude already knows). Consolidated reference file listings from 5 separate locations into a single "Load Documentation" section. Kept agent-centric design principles (MCP-specific value) and the critical warning about long-running server processes.
skill-check: Removed the "Testing this skill" section (meta-validation that doesn't help Claude perform the core task). Trimmed commands reference to essential commands. Added specific concrete actions to description (frontmatter schema, required fields, formatting issues, quality scoring with line numbers).
Want your remaining skills optimised too? 🚀
This PR covers 5 of your 8 skills to keep the contribution focused and reviewable. We have tooling that can go further:
SKILL.mdchanged in future PRs. Review mode works with zero secrets; maintainers can addTESSL_API_TOKENfor AI suggestions and the/apply-optimizecomment flow.Interested? Just tick the box below and we'll raise a follow-up PR:
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 and ask it to optimize your skill. Ping me - @rohan-tessl - if you hit any snags.
Thanks in advance 🙏