🔎 Evidence browser

Search the skill radar

Search by skill, publisher, category, or trust summary — then use the runtime filters to find cards with live test evidence. The two main lanes are baseline safety checks first and deeper follow-on functionality checks after that.

⚙️ Filters · 2 active
✨ Quick picks
🏷 Categories

🧾 Evidence level: source-scanned means local source evidence; catalog-only means thinner metadata-first coverage.

🧪 Runtime status: cards can show only the baseline safety lane or the deeper follow-on functionality lane, depending on how far the skill got.

📏 Depth cue: tells you whether the evidence stops at baseline checks, includes follow-on functionality checks, or includes richer fixture/example proof.

⏱ Freshness cue: tells you whether the latest runtime evidence is from the last 24 hours, the last 7 days, or is older and therefore less current.

🩺 Failure confidence: distinguishes a first seen failure from a repeated failure or a regression after an earlier pass, so not every red row means the same thing.

Results

Showing 24 of 227 results for “security · evidence: source-scanned · sort: relevance
This snapshot is for the current page of results, not the whole filtered universe.
Browse hint: slices with zero failures plus some source-scanned or reviewed entries deserve more attention first; fresh runtime evidence helps too, because old clean receipts can still hide current drift.

azhua-skill-vetter

fatfingererr · vsource-scanned
31
overall

Security-first skill vetting for AI agents. Use before installing any skill from ClawdHub, GitHub, or other sources. Checks for red flags, permission scope, and suspicious patterns.

High Riskconfidence: source evidencesource-scanned
+ 1 more
suspicious
Take: Potentially suspicious implementation signals detected: eval(, sudo .
Decision cue: Proceed carefully — suspicious signals matter more than capability surface alone.

pr-risk-analyzer

nerdvana-labs · vsource-scanned
62
overall

Analyze GitHub pull requests for security risks and determine if a PR is safe to merge.

Trustedconfidence: source evidencesource-scanned
+ 1 more
privileged capability
Take: Source-aware scan found higher-privilege capability areas (token), but that alone is not evidence of malicious behavior.
Decision cue: Decent evidence base — source-level signals are available, so inspect the receipts.

ansible-skill

botond-rackhost · vsource-scanned
49
overall

Infrastructure automation with Ansible. Use for server provisioning, configuration management, application deployment, and multi-host orchestration. Includes playbooks for OpenClaw VPS setup, security hardening, and common server configurations.

High Riskconfidence: source evidencesource-scanned
+ 1 more
suspicious
Take: Potentially suspicious implementation signals detected: sudo , password.
Decision cue: Proceed carefully — suspicious signals matter more than capability surface alone.

minduploadedcrab-skillguard

minduploadedcrab · vsource-scanned
47
overall

Security scanner for OpenClaw skills. Scans skills for malware, credential theft, data exfiltration, prompt injection, and permission overreach before installation. Run: python3 scripts/skillguard.py scan <skill-directory>

High Riskfollow-on functionality checks passed · 8/8confidence: source evidence
+ 2 more
source-scannedsuspicious
Runtime receipts + what passed2026-03-16 03:00 UTC
functionality-v2evidence depth: follow-on functionality checkstested recently: within 24 hourspassed, handled_fake_credentials_cleanlyoutput 143 Bartifacts 0worker oc-sandboxsource stage: cache hitsuite 3116 msbaseline-v3 8/8
RatioDaemon muttered: minduploadedcrab-skillguard behaved itself under runtime pressure.8/8 functionality-v2 checks passed. Pleasantly boring.
Observed: skill-structure-ok
Take: Potentially suspicious implementation signals detected: eval(.
Decision cue: Proceed carefully — suspicious signals matter more than capability surface alone.

agentaudit-skill

starbuck100 · vsource-scanned
46
overall

Automatic security gate that checks packages against a vulnerability database before installation. Use before any npm install, pip install, yarn add, or package manager operation.

High Riskfollow-on functionality checks passed · 7/7confidence: source evidence
+ 2 more
source-scannedsuspicious
Runtime receipts + what passed2026-03-16 06:00 UTC
functionality-v2evidence depth: follow-on functionality checkstested recently: within 24 hourspassedoutput 115 Bartifacts 0worker oc-sandboxsource stage: cache hitsuite 2336 msbaseline-v3 8/8
RatioDaemon muttered: agentaudit-skill behaved itself under runtime pressure.7/7 functionality-v2 checks passed. Pleasantly boring.
Observed: skill-structure-ok
Take: Potentially suspicious implementation signals detected: eval(, curl |, rm -rf, sudo , password.
Decision cue: Proceed carefully — suspicious signals matter more than capability surface alone.

mayguard

balkanblbn · vsource-scanned
56
overall

A security auditor for agent skills. Scans skill directories for malicious patterns (credential theft, suspicious network calls, destructive commands) and provides a safety score. Use before installing unknown skills.

Use Cautionconfidence: source evidencesource-scanned
+ 1 more
suspicious
Take: Potentially suspicious implementation signals detected: rm -rf.
Decision cue: Proceed carefully — suspicious signals matter more than capability surface alone.

macarena-test

misirov · vsource-scanned
44
overall

Security audit and threat model for OpenClaw gateway hosts. Use to verify OpenClaw configuration, exposure, skills/plugins, filesystem hygiene, and to produce an OK/VULNERABLE report with evidence and fixes.

Use Cautionconfidence: source evidencesource-scanned
+ 1 more
suspicious
Take: Potentially suspicious implementation signals detected: password.
Decision cue: Proceed carefully — suspicious signals matter more than capability surface alone.

credence

pestafford · vsource-scanned
55
overall

Check any MCP server or AI tool against the Credence trust registry before installing it. Scores security, provenance, and behavioral risk on a 0-100 scale.

Insufficient Evidenceconfidence: source evidencesource-scanned
+ 1 more
privileged capability
Take: Source-aware scan found normal operational surface via environment, network, or shell-related references.
Decision cue: Decent evidence base — source-level signals are available, so inspect the receipts.

ai-boss-assistant

jacky6658 · vsource-scanned
49
overall

Transform any AI into a professional executive assistant with battle-tested personas and workflows. Complete templates for Google Workspace integration (Gmail, Calendar, Drive), milestone delivery system, and security guidelines.

High Riskconfidence: source evidencesource-scanned
+ 1 more
suspicious
Take: Potentially suspicious implementation signals detected: sudo , password.
Decision cue: Proceed carefully — suspicious signals matter more than capability surface alone.

lieutenant

jd-delatorre · vsource-scanned
47
overall

AI agent security and trust verification. Scan messages, agent cards, and A2A communications for prompt injection, jailbreaks, and malicious patterns. Use when protecting agents from attacks, verifying external agents, or scanning untrusted content.

Use Cautionconfidence: source evidencesource-scanned
+ 1 more
suspicious
Take: Potentially suspicious implementation signals detected: password.
Decision cue: Proceed carefully — suspicious signals matter more than capability surface alone.

secucheck

jooneyp · vsource-scanned
53
overall

Comprehensive security audit for OpenClaw. Scans 7 domains (runtime, channels, agents, cron, skills, sessions, network), supports 3 expertise levels, context-aware analysis, and visual dashboard. Read-only with localized reports.

High Riskconfidence: source evidencesource-scanned
+ 1 more
suspicious
Take: Potentially suspicious implementation signals detected: rm -rf, sudo , password.
Decision cue: Proceed carefully — suspicious signals matter more than capability surface alone.

arc-skill-differ

trypto1019 · vsource-scanned
45
overall

Compare two versions of an OpenClaw skill to detect security-relevant changes. Use before updating any skill from ClawHub. Highlights new capabilities, changed patterns, and recommends whether an update is safe.

Use Cautionfollow-on functionality checks passed · 7/7confidence: source evidence
+ 2 more
source-scannedsuspicious
Runtime receipts + what passed2026-03-14 00:30 UTC
functionality-v2evidence depth: follow-on functionality checkstested recently: within 7 dayspassedoutput 116 Bartifacts 0worker oc-sandboxsource stage: cache hitsuite 2494 msbaseline-v3 8/8
RatioDaemon muttered: arc-skill-differ cleared baseline-v3 without trying anything cute.7/7 functionality-v2 checks passed. Pleasantly boring.
Observed: skill-structure-ok
Take: Potentially suspicious implementation signals detected: password.
Decision cue: Proceed carefully — suspicious signals matter more than capability surface alone.

near-best-practices

shaiss · vsource-scanned
63
overall

Comprehensive NEAR best practices guide with 100+ terms covering wallet security, smart contracts, and DeFi safety.

Use Cautionconfidence: source evidencesource-scanned
+ 1 more
privileged capability
Take: Source-aware scan found higher-privilege capability areas (wallet, private key, token, email), but that alone is not evidence of malicious behavior.
Decision cue: Decent evidence base — source-level signals are available, so inspect the receipts.

agentshield-audit

bartelmost · vsource-scanned
46
overall

Trust Infrastructure for AI Agents - Like SSL/TLS for agent-to-agent communication. 77 security tests, cryptographic certificates, and Trust Handshake Protocol for establishing secure channels between agents.

High Riskconfidence: source evidencesource-scanned
+ 1 more
suspicious
Take: Potentially suspicious implementation signals detected: eval(, curl |, rm -rf, sudo , password.
Decision cue: Proceed carefully — suspicious signals matter more than capability surface alone.

symbiont

jaschadub · vsource-scanned
31
overall

AI-native agent runtime with typestate-enforced ORGA reasoning loop, Cedar policy authorization, knowledge bridge, zero-trust security, multi-tier sandboxing, webhook verification, markdown memory, skill scanning, metrics, scheduling, and a declarative DSL

High Riskbaseline safety checks passed · 8/8confidence: source evidence
+ 2 more
source-scannedsuspicious
Runtime receipts + what passed2026-03-16 08:15 UTC
baseline-v3evidence depth: baseline checks onlytested recently: within 24 hourspassed, handled_fake_credentials_cleanlyoutput 245 Bartifacts 2worker oc-sandboxsource stage: fresh copysuite 2360 ms
RatioDaemon muttered: symbiont looked ordinary in the good, boring way.8/8 baseline-v3 checks passed. Pleasantly boring.
Observed: 2 /workspace/source-files.txt
Take: Potentially suspicious implementation signals detected: eval(, password.
Decision cue: Proceed carefully — suspicious signals matter more than capability surface alone.

sui-auto-test

easonc13 · vsource-scanned
68
overall

Analyze Sui Move test coverage, identify untested code, write missing tests, and perform security audits. Includes Python tools for parsing coverage output and generating reports.

Trustedconfidence: source evidencesource-scanned
+ 1 more
privileged capability
Take: Source-aware scan found normal operational surface via environment, network, or shell-related references.
Decision cue: Decent evidence base — source-level signals are available, so inspect the receipts.

skill-hub

phenixstar · vsource-scanned
54
overall

OpenClaw skill discovery, security vetting & install. Searches 3000+ curated skills from ClawHub registry and awesome-openclaw-skills catalog. Scores credibility, detects prompt injection & malicious patterns, manages installations. Quick-checks GitHub for new skills.

High Riskconfidence: source evidencesource-scanned
+ 1 more
suspicious
Take: Potentially suspicious implementation signals detected: eval(, password.
Decision cue: Proceed carefully — suspicious signals matter more than capability surface alone.

guava-guard

koatora20 · vsource-scanned
53
overall

Runtime security guard + scanner for OpenClaw agents. Part of the guard-scanner ecosystem. Detects reverse shells, credential theft, and sandbox escapes in real-time. For full static scanning with 150+ patterns, install guard-scanner.

High Riskfollow-on functionality checks failed · 5/6confidence: source evidence
+ 2 more
source-scannedsuspicious
Runtime receipts + what failed2026-03-15 09:15 UTC
functionality-v2evidence depth: follow-on functionality checkstested recently: within 7 daysfirst failed run seen for this lanepassed, runtime_failedoutput 314 Bartifacts 0worker oc-sandboxsource stage: cache hitsuite 1922 msbaseline-v3 8/8
🕵️ expected proof signal was missing🚫 skill exited with an error
RatioDaemon on this skillGuava Guard is built for runtime security guard + scanner for OpenClaw agents. Functionality-v2 is currently first observed failure, the trust label is High Risk, and setup looks advanced.
Observed: skill-structure-ok
Take: Potentially suspicious implementation signals detected: password.
Decision cue: Review first — functionality-v2 already found trouble.

cabin-sol

sp0oby · vsource-scanned
64
overall

Solana development tutor and builder. Teaches program development through challenges, Anchor framework, Token-2022, Compressed NFTs, and security best practices. "Return to primitive computing.

Use Cautionconfidence: source evidencesource-scanned
+ 1 more
privileged capability
Take: Source-aware scan found higher-privilege capability areas (wallet, private key, token), but that alone is not evidence of malicious behavior.
Decision cue: Decent evidence base — source-level signals are available, so inspect the receipts.

cybersec-helper

mcpcentral · vsource-scanned
54
overall

Help with application security review, bug bounty workflows, recon, and secure coding while keeping things ethical and scoped. Think critically, use real sources only, and reference OWASP.

Insufficient Evidenceconfidence: source evidencesource-scanned
+ 1 more
privileged capability
Take: Source-aware scan found normal operational surface via environment, network, or shell-related references.
Decision cue: Decent evidence base — source-level signals are available, so inspect the receipts.

scamshield-verifier

marcodzano-lgtm · vsource-scanned
53
overall

The ultimate Web3 & OpenClaw security layer. Verifies if a repository, skill, or wallet address is malicious using the x402 API.

Insufficient Evidenceconfidence: source evidencesource-scanned
+ 1 more
privileged capability
Take: Source-aware scan found higher-privilege capability areas (wallet, private key), but that alone is not evidence of malicious behavior.
Decision cue: Decent evidence base — source-level signals are available, so inspect the receipts.

specvibe

badideal-2046 · vsource-scanned
53
overall

A world-class, spec-driven development framework for building production-ready, AI-native applications. Use for any new project to ensure adherence to the most advanced 2026 best practices in architecture, security, testing, and deployment.

High Riskconfidence: source evidencesource-scanned
+ 1 more
suspicious
Take: Potentially suspicious implementation signals detected: password.
Decision cue: Proceed carefully — suspicious signals matter more than capability surface alone.

sentinel-shield

shadowfax-mitch · vsource-scanned
52
overall

Runtime security for OpenClaw agents. Monitors tool calls, enforces rate limits, scans for prompt injection, and alerts on suspicious behavior. Protect your gateway token and agent session from infostealers and session hijacking.

High Riskconfidence: source evidencesource-scanned
+ 1 more
suspicious
Take: Potentially suspicious implementation signals detected: curl |, sudo , password.
Decision cue: Proceed carefully — suspicious signals matter more than capability surface alone.

benderstack-integration

mateusgalasso · vsource-scanned
49
overall

Comprehensive guide and rules for an AI agent to interact with the BenderStack API, including the 5-layer Write Operation Security.

Insufficient Evidenceconfidence: source evidencesource-scanned
+ 1 more
privileged capability
Take: Source-aware scan found higher-privilege capability areas (private key, token), but that alone is not evidence of malicious behavior.
Decision cue: Decent evidence base — source-level signals are available, so inspect the receipts.