记忆

credential-exposure-map

试用

Map all credentials your OpenClaw agent can access. Scans env vars, config, memory, skills, MCP servers, git history. Generates exposure report with risk scoring. Activate when user says "credential audit", "exposure map", "security audit", or "what can my agent access".

它能做什么

Map all credentials your OpenClaw agent can access. Scans env vars, config, memory, skills, MCP servers, git history. Generates exposure report with risk scoring. Activate when user says "credential audit", "exposure map", "security audit", or "what can my agent access".

技能文档

Credential Exposure Map

See every credential your agent can touch. Score the blast radius. Shrink the attack surface.

When to Activate

  • User says: "credential audit", "exposure map", "security audit"
  • User says: "what can my agent access", "credential scan"
  • User says: "credential cleanup", "redact credentials"

Do NOT activate for normal work conversations.

Audit Flow

When user asks for a credential audit or exposure map:

  1. Run the scanner:
python3 skills/credential-exposure-map/scripts/scan_exposure.py

  1. Present the output in a clean format:

    • Summary: total findings, risk distribution
    • Credential Inventory: table with preview, risk level, source, location
    • Skill Capability Matrix: which skills have exec/read/network/write
    • Top recommendations for reducing exposure

  2. Highlight CRITICAL findings prominently.

Recommendation Generation

After presenting the scan, offer specific recommendations:

  • CRITICAL credentials in env: "Move to secrets vault or remove from agent context"
  • Credentials in MEMORY.md: "These persist across sessions. Consider redacting."
  • Skills with full exec+network: "Review if this skill needs both capabilities"
  • Secrets in git history: "Use git filter-branch or BFG to scrub history"
  • MCP servers with auth: "Verify each MCP server's permission scope"

Output Format

Summary

Scan Complete: X finding(s)
Risk: N Critical | N High | N Medium | N Low

Credential Inventory

Table showing each credential with:

  • Preview (first 8 chars + ***)
  • Risk level (CRITICAL/HIGH/MEDIUM/LOW)
  • Source (env/config/memory/skill/mcp/git)
  • Location (file:line or config path)

Skill Capability Matrix

Table showing each installed skill with exec/read/network/write permissions and risk score.

Security Considerations

  • NEVER output full credential values. Always mask as first8chars***
  • Report is saved to ~/.openclaw/credential-exposure-report.json with 600 permissions
  • Scan is read-only, does not modify any files
  • Skill capability inference is conservative (assumes exec = full filesystem access)

Detection Coverage

SourceWhat It Finds
Environment variablesAPI keys, tokens, passwords in process.env
openclaw.jsonProvider keys, webhook secrets, MCP auth
.env filesAny .env in workspace or ~/.openclaw
MEMORY.md / memory/*.mdCredentials written into agent memory
memory/*.jsonCredential caches, state files with tokens
Installed skillsSkills with exec, read, or network capabilities
MCP server configConnected services with auth scope
Git historySecrets committed in last 50 commits

相关技能

Skill Auditor

Automated security audit for AI agent skills. Use BEFORE installing any skill from ClawHub, GitHub, or other sources. Scans SKILL.md + all files for 30+ red...

作者 Zoran2 次安装

Clawditor

Audit an OpenClaw agent workspace and generate standardized evaluation reports, scores, and patches. Use when asked to review memory quality, retrieval effic...

32 次安装

Skill Auditor

Deep audit for installed ClawHub skills — usage analysis, permission review, conflict detection

4 次安装

Agentsec

按 OWASP Agentic Skills Top 10 审计已安装的 AI Agent 技能，输出文本、JSON、SARIF 或 HTML 报告。

29 次安装1 星标

OpenClaw Shield Quick Scan

Run a fast OpenClaw Shield scan on a folder or file, then summarize severity counts, top findings, and recommended next actions.

10 次安装

agent-canary

Plant decoy credentials in your OpenClaw workspace to detect malicious skill behavior. Canary tokens trigger alerts when read, copied, or exfiltrated. Activate when user says "deploy canary", "set up honeypot", "plant decoy", "canary check", "canary status", or "canary cleanup".

1 次安装