Skip to content
Harness Guide — Skill Recommendation logo

Harness Guide — Skill Recommendation

--- description: Triggered when the user says "recommend skill", "which skill", "harness guide", "skill recommendation", "what skill should I use", or "suggest a skill". Matches the best Skill for the user's scenario, shows installation status and invocation method. ---

SKILL.md

Full skill instructions

Harness Guide — Skill Recommendation


description: Triggered when the user says "recommend skill", "which skill", "harness guide", "skill recommendation", "what skill should I use", or "suggest a skill". Matches the best Skill for the user's scenario, shows installation status and invocation method.

Behavior

  1. Read references/​skill-guide.md for the scenario → Skill recommendation matrix
  2. Scan ~/​.claude/​skills/ for installed Skill list
  3. If user describes a specific scenario → precise match recommendation
  4. If no scenario described → display full recommendation matrix + decision tree

Matching Logic

Scenario Keywords → Skill Mapping

security audit / code audit / vulnerability scan
  → Check if security-review-skill-for-<project> exists
    → Yes → "Use your project-specific audit Skill directly"
    → No → "Recommend generating one with security-review-skill-creator first"

Docker / container / Dockerfile
  → security-review-skill-for-docker

Terraform / IaC / infrastructure as code
  → security-review-skill-for-terraform

SCA / vulnerability denoising / dependency vulnerabilities / CVE
  → sca-ai-denoise

supply chain / dependency poisoning / typosquatting
  → supply-chain-audit

Skill audit / third-party Skill / Skill security
  → skills-audit

web vulnerability / pentest / XSS / SQLi
  → web-vuln-analyzer (note: requires Docker configuration)

Android / APK / reverse engineering / frida
  → android-vuln-analyzer (note: requires toolchain)

new feature / feature development
  → superpowers (brainstorming → TDD)

bug / debug / debugging
  → superpowers (systematic-debugging)

code review
  → superpowers (requesting-code-review)

plan / task breakdown / planning
  → planning-with-files

create Skill / codify workflow
  → skill-creator

capture experience / lessons learned / knowledge extraction
  → claudeception

frontend / UI / component
  → frontend-design

design / poster / chart
  → canvas-design

Output Format

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🧭 Skill Recommendation
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Scenario: "your described scenario"

Recommended Skill:
  ✅ skill-name — description
     Invoke: "trigger keyword or command"

  Alternatives:
  ✅ alt-skill — description

Status legend:
  ✅ = Installed  ❌ = Not installed  ⚠️ = Needs configuration

If the recommended Skill is not installed:

❌ skill-name is not installed

Installation:
  Option A (Bundled): ln -sf ~/​.claude/​skills/​harness-en/​bundled-skills/​skill-name ~/​.claude/​skills/​skill-name
  Option B (Git):     cd ~/​.claude/​skills/ && git clone <url> skill-name

No Scenario: Display Full Guide

Read references/​skill-guide.md for the 4 categories + decision tree, combined with installation status.