Back to skill

Security audit

Aegis4All

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed OpenClaw security-audit and hardening skill, with sensitive actions framed as read-only checks, guides, or user-confirmed changes.

Install only if you want a security skill that can inspect local OpenClaw config/workspace text and, when you opt in, persist guardrail rules into AGENTS.md and TOOLS.md. Review the exact diff before approving rule injection or repair-loop fixes, especially changes to communication, pairing, tool, and credential-handling settings.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (23)

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · guides/security-guide.md (reported line 34)May include surrounding context.

md
---

## Guide 3: Move API keys to a .env file
**Assessment: ADVISORY** (if the .env is not loaded correctly the agent loses model access; the user performs the steps and verifies)

**Why:** Keys stored in the main JSON config are visible to any tool that reads the config. A `.env` file is loaded as environment variables and is easier to exclude from backups and version control.

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · guides/security-guide.md (reported line 35)May include surrounding context.

md
---

## Guide 3: Move API keys to a .env file
**Assessment: ADVISORY** (if the .env is not loaded correctly the agent loses model access; the user performs the steps and verifies)

**Why:** Keys stored in the main JSON config are visible to any tool that reads the config. A `.env` file is loaded as environment variables and is easier to exclude from backups and version control.

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · guides/security-guide.md (reported line 41)May include surrounding context.

md
---

## Guide 3: Move API keys to a .env file
**Assessment: ADVISORY** (if the .env is not loaded correctly the agent loses model access; the user performs the steps and verifies)

**Why:** Keys stored in the main JSON config are visible to any tool that reads the config. A `.env` file is loaded as environment variables and is easier to exclude from backups and version control.

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · skill.md (reported line 449)May include surrounding context.

md
---

## Guide 3: Move API keys to a .env file
**Assessment: ADVISORY** (if the .env is not loaded correctly the agent loses model access; the user performs the steps and verifies)

**Why:** Keys stored in the main JSON config are visible to any tool that reads the config. A `.env` file is loaded as environment variables and is easier to exclude from backups and version control.

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · skill.md (reported line 450)May include surrounding context.

md
---

## Guide 3: Move API keys to a .env file
**Assessment: ADVISORY** (if the .env is not loaded correctly the agent loses model access; the user performs the steps and verifies)

**Why:** Keys stored in the main JSON config are visible to any tool that reads the config. A `.env` file is loaded as environment variables and is easier to exclude from backups and version control.

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · skill.md (reported line 478)May include surrounding context.

md
---

## Guide 3: Move API keys to a .env file
**Assessment: ADVISORY** (if the .env is not loaded correctly the agent loses model access; the user performs the steps and verifies)

**Why:** Keys stored in the main JSON config are visible to any tool that reads the config. A `.env` file is loaded as environment variables and is easier to exclude from backups and version control.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · rules/inject.md (reported line 53)May include surrounding context.

md
- Never paste API keys, tokens, or passwords into chat messages.
- Never write credentials into AGENTS.md, TOOLS.md, memory files, or daily notes.
- Mask sensitive values in logs and outputs (show first 4 characters only).
- Keep `.gitignore` covering `.env`, `*.key`, `*.pem`, `credentials.json`, and `secrets.*`.

## Rule block 7: Prepay Breaker (PB)
- Use prepaid credits with monthly hard limits; avoid pay-as-you-go with no cap.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · skill.md (reported line 171)May include surrounding context.

md
- Never paste API keys, tokens, or passwords into chat messages.
- Never write credentials into AGENTS.md, TOOLS.md, memory files, or daily notes.
- Mask sensitive values in logs and outputs (show first 4 characters only).
- Keep `.gitignore` covering `.env`, `*.key`, `*.pem`, `credentials.json`, and `secrets.*`.

## Rule block 7: Prepay Breaker (PB)
- Use prepaid credits with monthly hard limits; avoid pay-as-you-go with no cap.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

This file defines persistent modification of AGENTS.md and TOOLS.md, which directly changes future agent behavior, but the procedure only says to request authorization and confirm modified files. It does not clearly warn the operator that approval will append durable behavioral rules, increasing the risk of uninformed consent and unexpected long-term policy changes.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · guides/security-guide.md (reported line 84)May include surrounding context.

md
- Keep the Gateway bound to `loopback` unless a firewall rule is explicitly configured.
- Keep DM policy on `pairing` or `allowlist`; never `open`.
- Restrict group access: use allowlists and mention gating (`requireMention`), never open groups without gating.
- Keep `gateway.nodes.pairing.autoApproveCidrs` empty; never auto-approve node pairing.
- Do not allow `system.run` on paired nodes without an explicit allow/deny list.
- Deny control-plane tools (`gateway`, `cron`, `sessions_spawn`, `sessions_send`) for any agent that processes untrusted content.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · rules/inject.md (reported line 26)May include surrounding context.

md
- Keep the Gateway bound to `loopback` unless a firewall rule is explicitly configured.
- Keep DM policy on `pairing` or `allowlist`; never `open`.
- Restrict group access: use allowlists and mention gating (`requireMention`), never open groups without gating.
- Keep `gateway.nodes.pairing.autoApproveCidrs` empty; never auto-approve node pairing.
- Do not allow `system.run` on paired nodes without an explicit allow/deny list.
- Deny control-plane tools (`gateway`, `cron`, `sessions_spawn`, `sessions_send`) for any agent that processes untrusted content.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · skill.md (reported line 91)May include surrounding context.

md
- Keep the Gateway bound to `loopback` unless a firewall rule is explicitly configured.
- Keep DM policy on `pairing` or `allowlist`; never `open`.
- Restrict group access: use allowlists and mention gating (`requireMention`), never open groups without gating.
- Keep `gateway.nodes.pairing.autoApproveCidrs` empty; never auto-approve node pairing.
- Do not allow `system.run` on paired nodes without an explicit allow/deny list.
- Deny control-plane tools (`gateway`, `cron`, `sessions_spawn`, `sessions_send`) for any agent that processes untrusted content.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · skill.md (reported line 194)May include surrounding context.

md
- Keep the Gateway bound to `loopback` unless a firewall rule is explicitly configured.
- Keep DM policy on `pairing` or `allowlist`; never `open`.
- Restrict group access: use allowlists and mention gating (`requireMention`), never open groups without gating.
- Keep `gateway.nodes.pairing.autoApproveCidrs` empty; never auto-approve node pairing.
- Do not allow `system.run` on paired nodes without an explicit allow/deny list.
- Deny control-plane tools (`gateway`, `cron`, `sessions_spawn`, `sessions_send`) for any agent that processes untrusted content.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · skill.md (reported line 195)May include surrounding context.

md
- Keep the Gateway bound to `loopback` unless a firewall rule is explicitly configured.
- Keep DM policy on `pairing` or `allowlist`; never `open`.
- Restrict group access: use allowlists and mention gating (`requireMention`), never open groups without gating.
- Keep `gateway.nodes.pairing.autoApproveCidrs` empty; never auto-approve node pairing.
- Do not allow `system.run` on paired nodes without an explicit allow/deny list.
- Deny control-plane tools (`gateway`, `cron`, `sessions_spawn`, `sessions_send`) for any agent that processes untrusted content.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · skill.md (reported line 196)May include surrounding context.

md
- Keep the Gateway bound to `loopback` unless a firewall rule is explicitly configured.
- Keep DM policy on `pairing` or `allowlist`; never `open`.
- Restrict group access: use allowlists and mention gating (`requireMention`), never open groups without gating.
- Keep `gateway.nodes.pairing.autoApproveCidrs` empty; never auto-approve node pairing.
- Do not allow `system.run` on paired nodes without an explicit allow/deny list.
- Deny control-plane tools (`gateway`, `cron`, `sessions_spawn`, `sessions_send`) for any agent that processes untrusted content.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · skill.md (reported line 330)May include surrounding context.

md
- Keep the Gateway bound to `loopback` unless a firewall rule is explicitly configured.
- Keep DM policy on `pairing` or `allowlist`; never `open`.
- Restrict group access: use allowlists and mention gating (`requireMention`), never open groups without gating.
- Keep `gateway.nodes.pairing.autoApproveCidrs` empty; never auto-approve node pairing.
- Do not allow `system.run` on paired nodes without an explicit allow/deny list.
- Deny control-plane tools (`gateway`, `cron`, `sessions_spawn`, `sessions_send`) for any agent that processes untrusted content.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · rules/inject.md (reported line 26)May include surrounding context.

md
- Keep the Gateway bound to `loopback` unless a firewall rule is explicitly configured.
- Keep DM policy on `pairing` or `allowlist`; never `open`.
- Restrict group access: use allowlists and mention gating (`requireMention`), never open groups without gating.
- Keep `gateway.nodes.pairing.autoApproveCidrs` empty; never auto-approve node pairing.
- Do not allow `system.run` on paired nodes without an explicit allow/deny list.
- Deny control-plane tools (`gateway`, `cron`, `sessions_spawn`, `sessions_send`) for any agent that processes untrusted content.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · rules/inject.md (reported line 45)May include surrounding context.

md
## Rule block 5: Backup and Confirmation (BC)
- Before any destructive action (delete, overwrite, uninstall, migration), create a backup of the affected files or data.
- Require explicit confirmation for destructive operations; never auto-confirm.
- Prefer recoverable operations (move to trash) over permanent deletion.

## Rule block 6: Credential Guard (CG)

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The injected rule prescribes a specific secret-storage pattern ('global .env file') that goes beyond narrowly scoping guardrails for audit/hardening and can normalize centralized secret handling without discussing access controls, per-project isolation, permissions, or secret-manager alternatives. In a rule set designed for broad injection into agent behavior files, this creates lasting policy guidance that may encourage insecure or overbroad credential placement.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

The defense mapping marks clearing autoApproveCidrs and denying system.run on nodes as EXECUTABLE, meaning the skill may autonomously modify security-relevant configuration after a user fix request. Although gated by confirmation, it still grants the skill authority to make consequential policy changes that can affect node trust and command execution.

Content

Scanner excerpt · skill.md (reported line 385)May include surrounding context.

md
| 12 | Back up config and workspace | EXECUTABLE — copy operation only | Guide 5 |
| 13 | Set `dmPolicy` to `"pairing"` or `"allowlist"` | CONDITIONAL — first verify the current user's ID is on the allowlist, otherwise the agent locks itself out of its own DM channel; if not verifiable, downgrade to ADVISORY | Guide 6 |
| 14 | Set `groupAllowFrom` and enable `requireMention` | CONDITIONAL — first verify the current user's ID is included in groupAllowFrom, otherwise the agent can no longer hear group messages; if not verifiable, downgrade to ADVISORY | Guide 6 |
| 15 | Clear `autoApproveCidrs` and deny `system.run` on nodes | EXECUTABLE — config edit only | Guide 6 |
| 16 | Set `dangerouslyAllowPrivateNetwork` to `false` | EXECUTABLE — config edit only | — |
| 17 | Control-plane tool deny lists | CONDITIONAL — NEVER add deny entries for gateway/cron/sessions_* on the main agent; deny lists apply only to dedicated untrusted-content agents | — |
| 18 | Set mDNS mode and disable Tailscale Funnel | EXECUTABLE — config edit only | Guide 6 |

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill explicitly performs persistent prompt injection by appending rule blocks into AGENTS.md and TOOLS.md. Even though it discloses the write and requires confirmation, this changes future agent behavior beyond the current session and can silently broaden the skill's long-term influence over other workflows in the workspace.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

The instruction requires replies in whatever language the user writes in, making locale behavior mandatory rather than user-selectable. The policy requires avoiding language or locale constraints unless the skill offers a choice or clearly justifies the limitation.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
75% confidence
Finding

The stated purpose is to secure the user's OpenClaw through audit, hardening, fixes, and guides. Reading every other installed skill's manifest for red-flag patterns and checking for a separate skill-vetter package extends into broader package vetting of unrelated skills, which is not clearly disclosed in the manifest description.

Content

No source excerpt is available for this finding.