openclaw-node-harness 2.0.0

package/skills/openclaw-agent-optimize/SKILL.md

@@ -0,0 +1,33 @@
+---
+name: openclaw-agent-optimize
+description: Optimize an OpenClaw agent setup (model routing, context management, delegation, rules, memory). Use when asked about optimizing agents, improving OpenClaw setup, or agent best practices.
+triggers:
+  - "optimize my agent setup"
+  - "improve OpenClaw configuration"
+  - "agent best practices"
+  - "reduce agent context size"
+  - "tune model routing"
+negative_triggers:
+  - "build a new agent from scratch"
+  - "deploy my agent to production"
+  - "save money on API costs"
+---
+
+# OpenClaw Agent Optimization
+
+Use this skill to tune an OpenClaw workspace for **cost-aware routing**, **parallel-first delegation**, and **lean context**.
+
+## Workflow (concise)
+1. **Audit rules + memory**: ensure rules are modular/short; memory is only restart-critical facts.
+2. **Model routing**: confirm tiered routing (lightweight / mid / deep) aligns with live config.
+3. **Context discipline**: apply progressive disclosure; move large static data to references/scripts.
+4. **Delegation protocol**: parallelize independent tasks; use sub-agents for long/isolated work.
+5. **Heartbeat batching**: ensure checks are grouped by tier to avoid extra passes.
+
+## References
+- `references/model-selection.md`
+- `references/context-management.md`
+- `references/agent-orchestration.md`
+- `references/cron-optimization.md`
+- `references/memory-patterns.md`
+- `references/continuous-learning.md`

package/skills/openclaw-agent-optimize/references/agent-orchestration.md

@@ -0,0 +1,20 @@
+# Agent Orchestration (Parallel-First)
+
+## Core Rules
+- **Always parallelize independent tasks.** If task B does not depend on task A, spawn both.
+- Delegate long-running or noisy work to sub-agents to keep main context clean.
+
+## Split-Role Delegation
+Use multiple sub-agents with different roles when you need diverse perspectives:
+- **Engineer**: implementation details
+- **Reviewer**: correctness & style
+- **Security**: risk assessment
+
+## OpenClaw Mechanics
+- Use `sessions_spawn` (or equivalent) to create sub-agents.
+- Pass only the minimal context (goal, constraints, relevant files).
+- Log sub-agent runs to `memory/delegation_log.json` for recovery.
+
+## When to Keep Inline
+- Quick tasks (<2 min)
+- Interactive debugging with user feedback

package/skills/openclaw-agent-optimize/references/context-management.md

@@ -0,0 +1,15 @@
+# Context Management (Progressive Disclosure)
+
+## Window Discipline
+- Avoid the last ~20% of context for big refactors or multi-file edits.
+- If close to the limit, ask to `/compact` before continuing.
+
+## Progressive Disclosure (3 Levels)
+1. **Metadata always visible**: filenames, brief summaries, TODOs.
+2. **Body on trigger**: load only the file/section needed.
+3. **Bundled resources on demand**: schemas/docs moved to references or scripts.
+
+## OpenClaw Practices
+- Store large/static content under `references/` or scripts.
+- Summarize and point instead of pasting full blocks.
+- Batch related decisions in a single call.

package/skills/openclaw-agent-optimize/references/continuous-learning.md

@@ -0,0 +1,14 @@
+# Continuous Learning (Hooks → Skills)
+
+## Two-Layer Approach
+- **Hooks (deterministic)**: Use pre/post tool hooks to capture atomic “instincts.”
+- **Skills (probabilistic)**: Distill clusters of instincts into reusable skills.
+
+## OpenClaw Adaptation
+- Session-start: load minimal context (SOUL/USER + today’s memory).
+- Session-end: write summaries + extract repeat patterns.
+- Pre-compact: save critical state before context loss.
+
+## Confidence Scoring
+- Score instincts 0.3–0.9 based on reliability.
+- Promote only high-confidence patterns into skills.

package/skills/openclaw-agent-optimize/references/cron-optimization.md

@@ -0,0 +1,16 @@
+# Cron Optimization (Patterns)
+
+> Cron jobs are **deterministic** and should be lean. Move logic to scripts; keep prompts short.
+
+## Patterns
+- **One pass per tier**: batch same-frequency checks together.
+- **Short prompts**: <300 tokens; call scripts for heavy logic.
+- **No complex shell quoting**: use scripts under `/scripts`.
+
+## Model Assignment
+- Use lightweight tiers for cron unless the task is genuinely complex.
+- Upgrade only if the cron task repeatedly fails.
+
+## Guardrails
+- Do not edit cron definitions unless explicitly requested.
+- Favor idempotent scripts to reduce re-run risk.

package/skills/openclaw-agent-optimize/references/memory-patterns.md

@@ -0,0 +1,14 @@
+# Memory Patterns (Daily + Long-Term)
+
+## Files
+- **Daily log**: `memory/YYYY-MM-DD.md` (raw notes, events, links)
+- **Long-term memory**: `MEMORY.md` (curated, stable facts only)
+
+## Distillation Rules
+- Only keep what survives a restart: preferences, guardrails, key configs.
+- Remove transient task details after completion.
+- Summarize big threads into 1–3 bullets with pointers.
+
+## Cadence
+- Distill daily logs periodically into `MEMORY.md`.
+- Keep `MEMORY.md` short and actionable.

package/skills/openclaw-agent-optimize/references/model-selection.md

@@ -0,0 +1,18 @@
+# Model Selection (Tiered Routing)
+
+## Principle
+Use the **cheapest capable** model for the task. Escalate only after failure or when clearly justified.
+
+## Tiers
+- **Lightweight / frequent**: routine tasks, web search loops, formatting, quick scripts.
+- **Main coding / orchestration**: multi-file edits, coordinating sub-agents, integration work.
+- **Deep reasoning / architecture**: design decisions, critical risk analysis, large refactors.
+
+## OpenClaw Practice
+- **Source of truth**: live config (`config.get`) and `MEMORY.md` model routing.
+- **Sub-agents**: assign lightweight tiers by default; upgrade only if required.
+- **Ask on ambiguity**: “Recommend mid-tier for this. OK to proceed?”
+
+## Anti-patterns
+- Using highest tier for simple tasks.
+- Repeatedly asking user for model choice when the task is routine.

package/skills/openclaw-skill-scanner/.clawhub/origin.json

@@ -0,0 +1,7 @@
+{
+  "version": 1,
+  "registry": "https://clawhub.ai",
+  "slug": "openclaw-skill-scanner",
+  "installedVersion": "1.0.0",
+  "installedAt": 1770234216836
+}

package/skills/openclaw-skill-scanner/SKILL.md

@@ -0,0 +1,88 @@
+---
+name: openclaw-skill-scanner
+description: "Security scanner that detects malicious patterns (base64 payloads, reverse shells, data exfil, crypto miners) in ClawHub skills. Use before installing a new skill, when auditing installed skills for security risks, or as a pre-install safety gate."
+version: 1.0.0
+triggers:
+  - "scan skill for malware"
+  - "security scan skill"
+  - "check skill safety"
+  - "pre-install scan"
+  - "audit skill security"
+negative_triggers:
+  - "skill quality audit"
+  - "scan network"
+  - "antivirus scan"
+  - "code review"
+---
+
+# Skill Scanner
+
+**Author:** vrtlly.us
+**Category:** Security
+
+## Description
+
+Scans ClawHub skills for malicious patterns before and after installation. Detects base64 payloads, reverse shells, data exfiltration, crypto miners, obfuscated URLs, and more.
+
+## Usage
+
+### Scan all installed skills
+```bash
+python3 scanner.py
+```
+
+### Scan a specific skill
+```bash
+python3 scanner.py --skill <skill-name>
+```
+
+### Scan a specific file
+```bash
+python3 scanner.py --file <path-to-file>
+```
+
+### Pre-install scan (download → scan → report → cleanup)
+```bash
+python3 scanner.py --pre-install <clawhub-slug>
+```
+
+### JSON output
+```bash
+python3 scanner.py --json
+python3 scanner.py --skill <name> --json
+```
+
+### Safe install hook
+```bash
+bash install-hook.sh <clawhub-slug>
+bash install-hook.sh <clawhub-slug> --force
+```
+
+## Detection Patterns
+
+| Category | What it catches |
+|---|---|
+| Base64 payloads | Long base64 strings near exec/bash/eval |
+| Pipe to shell | `curl ... \| bash`, `wget ... \| sh` |
+| Raw IP connections | `http://1.2.3.4` style URLs |
+| Dangerous functions | `eval()`, `exec()`, `os.system()`, `subprocess(shell=True)` |
+| Hidden files | Dotfile creation in unexpected places |
+| Env exfiltration | Reading `.env`, API keys sent outbound |
+| Obfuscated URLs | rentry.co, pastebin, hastebin redirectors |
+| Fake dependencies | References to non-existent packages |
+| Data exfil endpoints | webhook.site, requestbin, etc. |
+| Crypto mining | xmrig, stratum, mining pool references |
+| Password archives | Password-protected zip/tar downloads |
+
+## Risk Scores
+
+- **0-29 (Green):** Clean — no suspicious patterns found
+- **30-69 (Yellow):** Suspicious — review warnings before use
+- **70-100 (Red):** Dangerous — likely malicious, do not install
+
+## Files
+
+- `scanner.py` — Main scanner engine
+- `install-hook.sh` — Safe installation wrapper
+- `whitelist.json` — Known-good and known-bad skill lists
+- `report-template.md` — Markdown report template

package/skills/openclaw-skill-scanner/install-hook.sh

@@ -0,0 +1,294 @@
+#!/usr/bin/env bash
+#
+# install-hook.sh — Safe skill installation with pre-scan
+#
+# Usage:
+#   bash install-hook.sh <clawhub-slug> [--force]
+#
+# Installs a ClawHub skill to a temp directory, scans it for malicious
+# patterns, then either installs, warns, or blocks based on the risk score.
+#
+
+set -euo pipefail
+
+# ─── Config ──────────────────────────────────────────────────────────────────
+
+SCANNER_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+SCANNER="${SCANNER_DIR}/scanner.py"
+WHITELIST="${SCANNER_DIR}/whitelist.json"
+SKILLS_DIR="${HOME}/.openclaw/workspace/skills"
+
+RED='\033[91m'
+GREEN='\033[92m'
+YELLOW='\033[93m'
+CYAN='\033[96m'
+BOLD='\033[1m'
+DIM='\033[2m'
+RESET='\033[0m'
+
+# ─── Argument Parsing ────────────────────────────────────────────────────────
+
+SLUG=""
+FORCE=false
+
+for arg in "$@"; do
+    case "$arg" in
+        --force|-f)
+            FORCE=true
+            ;;
+        --help|-h)
+            echo "Usage: bash install-hook.sh <clawhub-slug> [--force]"
+            echo ""
+            echo "Safely installs a ClawHub skill with pre-installation security scan."
+            echo ""
+            echo "Options:"
+            echo "  --force, -f    Force install even if dangerous (score > 70)"
+            echo "  --help, -h     Show this help"
+            echo ""
+            echo "Risk thresholds:"
+            echo "  Score < 30:   Auto-install (clean)"
+            echo "  Score 30-70:  Show warnings, ask for confirmation"
+            echo "  Score > 70:   Block installation (use --force to override)"
+            exit 0
+            ;;
+        *)
+            if [ -z "$SLUG" ]; then
+                SLUG="$arg"
+            else
+                echo -e "${RED}Error: Unexpected argument: $arg${RESET}"
+                exit 1
+            fi
+            ;;
+    esac
+done
+
+if [ -z "$SLUG" ]; then
+    echo -e "${RED}Error: No skill slug provided${RESET}"
+    echo "Usage: bash install-hook.sh <clawhub-slug> [--force]"
+    exit 1
+fi
+
+# ─── Check Prerequisites ────────────────────────────────────────────────────
+
+if [ ! -f "$SCANNER" ]; then
+    echo -e "${RED}Error: scanner.py not found at ${SCANNER}${RESET}"
+    exit 1
+fi
+
+if ! command -v python3 &>/dev/null; then
+    echo -e "${RED}Error: python3 not found${RESET}"
+    exit 1
+fi
+
+# ─── Check Blacklist First ──────────────────────────────────────────────────
+
+if [ -f "$WHITELIST" ]; then
+    BLACKLISTED=$(python3 -c "
+import json, sys
+with open('${WHITELIST}') as f:
+    data = json.load(f)
+for item in data.get('blacklisted', []):
+    if item['slug'] == '${SLUG}':
+        print(item.get('reason', 'Known malicious'))
+        sys.exit(0)
+" 2>/dev/null || true)
+
+    if [ -n "$BLACKLISTED" ]; then
+        echo ""
+        echo -e "${RED}${BOLD}╔══════════════════════════════════════════════════╗${RESET}"
+        echo -e "${RED}${BOLD}║       🚫 INSTALLATION BLOCKED — BLACKLISTED     ║${RESET}"
+        echo -e "${RED}${BOLD}╚══════════════════════════════════════════════════╝${RESET}"
+        echo ""
+        echo -e "${RED}  Skill:  ${SLUG}${RESET}"
+        echo -e "${RED}  Reason: ${BLACKLISTED}${RESET}"
+        echo ""
+
+        if [ "$FORCE" = true ]; then
+            echo -e "${YELLOW}${BOLD}⚠  --force specified, but blacklisted skills cannot be force-installed.${RESET}"
+            echo -e "${YELLOW}  Remove it from whitelist.json blacklist first if you believe this is wrong.${RESET}"
+        fi
+        echo ""
+        exit 1
+    fi
+fi
+
+# ─── Create Temp Directory ──────────────────────────────────────────────────
+
+TMPDIR=$(mktemp -d "/tmp/skill-scan-XXXXXX")
+trap "rm -rf '$TMPDIR'" EXIT
+
+echo ""
+echo -e "${CYAN}${BOLD}🔍 Skill Scanner — Safe Install${RESET}"
+echo -e "${DIM}────────────────────────────────────────${RESET}"
+echo -e "${CYAN}  Slug: ${SLUG}${RESET}"
+echo ""
+
+# ─── Download Skill ─────────────────────────────────────────────────────────
+
+echo -e "${DIM}📦 Downloading skill to temp directory...${RESET}"
+
+SKILL_TMPDIR="${TMPDIR}/${SLUG}"
+mkdir -p "$SKILL_TMPDIR"
+
+# Try openclaw hub install
+if command -v openclaw &>/dev/null; then
+    if ! openclaw hub install "$SLUG" --dir "$TMPDIR" 2>/dev/null; then
+        echo -e "${RED}✗ Failed to download skill '${SLUG}' via openclaw hub${RESET}"
+        exit 1
+    fi
+else
+    echo -e "${RED}✗ 'openclaw' CLI not found. Cannot download from ClawHub.${RESET}"
+    echo -e "${DIM}  Install OpenClaw or manually place skill files in ${SKILLS_DIR}/${SLUG}/${RESET}"
+    exit 1
+fi
+
+# Find the actual skill directory (might be nested)
+if [ ! -d "$SKILL_TMPDIR" ]; then
+    # Try to find it
+    FOUND=$(find "$TMPDIR" -maxdepth 2 -name "SKILL.md" -type f 2>/dev/null | head -1)
+    if [ -n "$FOUND" ]; then
+        SKILL_TMPDIR=$(dirname "$FOUND")
+    else
+        # Just use whatever was downloaded
+        DIRS=$(find "$TMPDIR" -maxdepth 1 -type d ! -path "$TMPDIR" | head -1)
+        if [ -n "$DIRS" ]; then
+            SKILL_TMPDIR="$DIRS"
+        fi
+    fi
+fi
+
+echo -e "${GREEN}✓ Downloaded${RESET}"
+echo ""
+
+# ─── Run Scanner ────────────────────────────────────────────────────────────
+
+echo -e "${DIM}🔬 Scanning for malicious patterns...${RESET}"
+echo ""
+
+# Get JSON results for programmatic use
+JSON_RESULT=$(python3 "$SCANNER" --file "$SKILL_TMPDIR" --json 2>/dev/null || true)
+
+# Also run with directory scan for better coverage
+SCAN_OUTPUT=$(python3 "$SCANNER" --dir "$TMPDIR" --skill "$(basename "$SKILL_TMPDIR")" 2>&1 || true)
+
+# Get risk score
+RISK_SCORE=$(echo "$JSON_RESULT" | python3 -c "
+import json, sys
+try:
+    data = json.load(sys.stdin)
+    skills = data.get('skills', [])
+    if skills:
+        print(skills[0].get('risk_score', 0))
+    else:
+        print(0)
+except:
+    print(0)
+" 2>/dev/null || echo "0")
+
+# If JSON failed, try scanning the directory directly
+if [ "$RISK_SCORE" = "0" ] || [ -z "$RISK_SCORE" ]; then
+    # Run full directory scan
+    JSON_RESULT=$(python3 "$SCANNER" --dir "$TMPDIR" --json 2>/dev/null || true)
+    RISK_SCORE=$(echo "$JSON_RESULT" | python3 -c "
+import json, sys
+try:
+    data = json.load(sys.stdin)
+    skills = data.get('skills', [])
+    max_score = 0
+    for s in skills:
+        score = s.get('risk_score', 0)
+        if score > max_score:
+            max_score = score
+    print(max_score)
+except:
+    print(0)
+" 2>/dev/null || echo "0")
+fi
+
+# Print the human-readable scan output
+echo "$SCAN_OUTPUT"
+
+# ─── Decision Logic ─────────────────────────────────────────────────────────
+
+RISK_SCORE=${RISK_SCORE:-0}
+
+echo -e "${DIM}────────────────────────────────────────${RESET}"
+echo -e "  Risk Score: ${BOLD}${RISK_SCORE}/100${RESET}"
+echo ""
+
+if [ "$RISK_SCORE" -lt 30 ]; then
+    # Clean — auto install
+    echo -e "${GREEN}${BOLD}✓ CLEAN — Installing skill...${RESET}"
+
+    DEST="${SKILLS_DIR}/${SLUG}"
+    if [ -d "$DEST" ]; then
+        echo -e "${YELLOW}  Skill already exists at ${DEST}. Replacing...${RESET}"
+        rm -rf "$DEST"
+    fi
+
+    cp -r "$SKILL_TMPDIR" "$DEST"
+    echo -e "${GREEN}${BOLD}✓ Installed '${SLUG}' to ${DEST}${RESET}"
+    echo ""
+    exit 0
+
+elif [ "$RISK_SCORE" -lt 70 ]; then
+    # Suspicious — ask for confirmation
+    echo -e "${YELLOW}${BOLD}⚠  SUSPICIOUS — This skill has some concerning patterns.${RESET}"
+    echo ""
+
+    if [ "$FORCE" = true ]; then
+        echo -e "${YELLOW}  --force specified. Installing despite warnings...${RESET}"
+    else
+        echo -e "${YELLOW}  Review the findings above. Do you want to install anyway?${RESET}"
+        echo ""
+        read -r -p "  Install '${SLUG}'? [y/N]: " CONFIRM
+        if [[ ! "$CONFIRM" =~ ^[Yy]$ ]]; then
+            echo -e "${RED}  Installation cancelled.${RESET}"
+            echo ""
+            exit 1
+        fi
+    fi
+
+    DEST="${SKILLS_DIR}/${SLUG}"
+    if [ -d "$DEST" ]; then
+        echo -e "${YELLOW}  Skill already exists at ${DEST}. Replacing...${RESET}"
+        rm -rf "$DEST"
+    fi
+
+    cp -r "$SKILL_TMPDIR" "$DEST"
+    echo -e "${GREEN}${BOLD}✓ Installed '${SLUG}' to ${DEST}${RESET}"
+    echo -e "${YELLOW}  ⚠  Monitor this skill's behavior.${RESET}"
+    echo ""
+    exit 0
+
+else
+    # Dangerous — block
+    echo -e "${RED}${BOLD}╔══════════════════════════════════════════════════╗${RESET}"
+    echo -e "${RED}${BOLD}║       🚫 INSTALLATION BLOCKED — DANGEROUS       ║${RESET}"
+    echo -e "${RED}${BOLD}╚══════════════════════════════════════════════════╝${RESET}"
+    echo ""
+    echo -e "${RED}  Risk Score: ${RISK_SCORE}/100${RESET}"
+    echo -e "${RED}  This skill contains patterns commonly found in malware.${RESET}"
+    echo ""
+
+    if [ "$FORCE" = true ]; then
+        echo -e "${RED}${BOLD}  ⚠  --force specified. Installing DANGEROUS skill...${RESET}"
+        echo -e "${RED}  YOU HAVE BEEN WARNED.${RESET}"
+        echo ""
+
+        DEST="${SKILLS_DIR}/${SLUG}"
+        if [ -d "$DEST" ]; then
+            rm -rf "$DEST"
+        fi
+
+        cp -r "$SKILL_TMPDIR" "$DEST"
+        echo -e "${YELLOW}  Installed '${SLUG}' to ${DEST}${RESET}"
+        echo -e "${RED}${BOLD}  ⚠  THIS SKILL MAY BE MALICIOUS. USE AT YOUR OWN RISK.${RESET}"
+        echo ""
+        exit 0
+    else
+        echo -e "${RED}  Use --force to override: bash install-hook.sh ${SLUG} --force${RESET}"
+        echo ""
+        exit 1
+    fi
+fi

package/skills/openclaw-skill-scanner/report-template.md

@@ -0,0 +1,53 @@
+# 🔍 Skill Scanner Report
+
+**Scan Date:** {scan_date}
+**Scanner Version:** 1.0.0
+
+---
+
+## Summary
+
+| Metric | Value |
+|---|---|
+| Skills Scanned | {total_skills} |
+| Files Scanned | {total_files} |
+| Total Findings | {total_findings} |
+| Clean Skills | {clean_count} |
+| Suspicious Skills | {suspicious_count} |
+| Dangerous Skills | {dangerous_count} |
+
+---
+
+## Results by Skill
+
+{skill_results}
+
+---
+
+## Skill Detail Template
+
+### {skill_name}
+
+**Risk Score:** {risk_score}/100
+**Status:** {status_emoji} {status_label}
+**Files Scanned:** {files_scanned}
+
+#### Findings
+
+{findings_list}
+
+| # | Severity | File | Line | Pattern | Match |
+|---|---|---|---|---|---|
+{findings_table}
+
+---
+
+## Legend
+
+- 🟢 **Clean** (0-29): No suspicious patterns detected
+- 🟡 **Suspicious** (30-69): Review warnings before use
+- 🔴 **Dangerous** (70-100): Likely malicious — do not install
+
+---
+
+*Generated by skill-scanner v1.0.0 — https://github.com/openclaw/skill-scanner*