Skip to content
Back to skills

Iknowkungfu

ASecurity

Skill discovery engine. Analyzes what your agent does and recommends ClawHub skills you're missing. Use when: /kungfu, /kungfu-scan, /kungfu-gaps, 'what skills am I missing', 'recommend skills', 'what should I install', 'skill discovery'.

  • 3 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added September 9, 2026
toolsrustgonodeawsapisecurity

Works with

  • api

Security analysis

A100/100

Scanned September 9, 2026

npx -y skills add Lord1Egypt/awesome-skill-forge --skill iknowkungfu --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Iknowkungfu?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Iknowkungfu
[![Security: A β€” Skills Directory](https://www.skillsdirectory.com/api/skills/lord1egypt-iknowkungfu/badge)](https://www.skillsdirectory.com/skills/lord1egypt-iknowkungfu)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: Iknowkungfu
description: "Skill discovery engine. Analyzes what your agent does and recommends ClawHub skills you're missing. Use when: /kungfu, /kungfu-scan, /kungfu-gaps, 'what skills am I missing', 'recommend skills', 'what should I install', 'skill discovery'."
---

# iknowkungfu πŸ₯‹

Skill discovery in 3 phases:
1. **Profile** πŸ” β€” Analyze your workflow (memory, skills, crons, logs)
2. **Match** 🎯 β€” Cross-reference against curated ClawHub index
3. **Recommend** πŸ“‹ β€” Prioritized suggestions with trust scores

100% local. No data leaves your machine.

## Commands

`/kungfu` full scan | `/kungfu-scan` profile only | `/kungfu-gaps` uncovered areas | `/kungfu-update` refresh index

## Phase 1: Profile

See `references/workflow-analysis.md` for full procedure.

Read these sources to build a Workflow Profile:
- **MEMORY.md + daily logs** β€” recurring topics, tools, domains
- **Installed skills** β€” list from BOTH `~/.openclaw/skills/` AND system paths (e.g. `/opt/homebrew/lib/node_modules/openclaw/skills/`). Check ALL install locations. Map to categories via `data/workflow-patterns.json`
- **AGENTS.md + config** β€” user role, tool preferences, model budget signals
- **HEARTBEAT.md + crons** β€” automated/scheduled responsibilities
- **Recent logs (7 days)** β€” dominant task types, frequent commands

Quick security check while reading skills: scan for base64, curl/wget, eval/exec, env var harvesting. Flag warnings. For deep scanning, recommend ClawSpa.

Output the Workflow Profile (template in references/workflow-analysis.md).

## Phase 2: Match

See `references/recommendation-engine.md` for full procedure.

Load `data/skills-catalogue.json`. For each gap in the profile:
1. Find matching skills by category
2. Score candidates (see `references/scoring.md`)
3. Filter already-installed skills (check ALL install paths: user, system, workspace)
4. Filter skills whose functionality is already covered by existing config (e.g. memoryFlush covers session wrap-up, gog covers Gmail)
5. Rank by score, deduplicate overlaps

## Phase 3: Validate Before Recommending

Before presenting, run each candidate through a relevance check:
- Does the user actually use this tool/service? (e.g. don't recommend Slack if they never mention it)
- Is equivalent functionality already covered by a system skill, config setting, or existing workflow?
- Would this realistically fit the user's setup? (solo builder vs team, macOS vs Linux, budget signals)

Drop candidates that fail. Better 2 genuinely useful than 5 with 3 irrelevant. If all fail: "gap detected but no relevant match for your setup."

## Phase 4: Recommend

Present top 5:

```
πŸ₯‹ I KNOW KUNG FU β€” Recommendations
═══════════════════════════════════════
1. 🟒 skill-name (β˜… 4.5)
   Category: [cat] | Author: [author]
   Why: [1-2 sentences tied to YOUR workflow]
   Install: clawhub install skill-name
   ─────────────────────────────────
[up to 5]
═══════════════════════════════════════
πŸ’‘ /kungfu-gaps for all uncovered areas
═══════════════════════════════════════
```

## Trust Scoring

See `references/scoring.md`. Factors: downloads (25%), stars (20%), author rep (15%), recency (15%), permissions (15%), security (10%). Never recommend: <50 downloads, VirusTotal flags, no author, excessive unjustified permissions.

## Safeguards

- READ-ONLY. Never installs, modifies, or removes anything. Zero network calls.
- Only recommends skills passing trust AND relevance thresholds.
- Honest about confidence. If no good match exists, says so.
- NEVER include full file contents in output. Only summarize patterns and categories.
- NEVER print API keys, tokens, passwords, SSH keys, or any credential-like strings found in any file.
- When reporting security flags, describe the PATTERN found (e.g. "env var reference in script"), never quote the actual value.
- Redact any file paths that contain usernames or home directories in output.

## Limitations

Catalogue is bundled (may lag). Trust scores are heuristic. <7 days history = less accurate.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

View all in tools β†’

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.