Security audit

Pad Mode

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed planning workflow that creates resumable plan files and can use sub-agents only after user approval.

Install only if you want a persistent planning workflow. Avoid putting secrets or sensitive business details in plan requests unless you are comfortable with those details being saved in OpenClaw plan files, and choose background/sub-agent execution only for tasks you have reviewed and approved.

A.I.G

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

SkillSpector

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
Rogue AgentSelf-Modification, Session Persistence
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (11)

Tool Parameter Abuse

High

Category: Tool Misuse
Confidence: 80% confidence
Finding: Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).
Content: Scanner excerpt · README.md (reported line 133)May include surrounding context.
md
- [x] **T1.1** Initialize Express project with TypeScript - Deliverable: Project scaffold with tsconfig, package.json, folder structure - Dependencies: none - [x] **T1.2** Implement CRUD endpoints (GET/POST/PUT/DELETE /todos) - Deliverable: 4 working endpoints with validation - Dependencies: T1.1 - [ ] **T2.1** Add JWT authentication

Tool Parameter Abuse

High

Category: Tool Misuse
Confidence: 80% confidence
Finding: Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).
Content: Scanner excerpt · README_zh.md (reported line 131)May include surrounding context.
md
- [x] **T1.1** 初始化 Express + TypeScript 项目 - 交付物: 项目脚手架，含 tsconfig、package.json、目录结构 - 依赖: 无 - [x] **T1.2** 实现 CRUD 接口（GET/POST/PUT/DELETE /todos） - 交付物: 4 个带验证的可用接口 - 依赖: T1.1 - [ ] **T2.1** 添加 JWT 认证

Tool Parameter Abuse

High

Category: Tool Misuse
Confidence: 80% confidence
Finding: Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).
Content: Scanner excerpt · README_zh.md (reported line 131)May include surrounding context.
md
- [x] **T1.1** 初始化 Express + TypeScript 项目 - 交付物: 项目脚手架,含 tsconfig、package.json、目录结构 - 依赖: 无 - [x] **T1.2** 实现 CRUD 接口(GET/POST/PUT/DELETE /todos) - 交付物: 4 个带验证的可用接口 - 依赖: T1.1 - [ ] **T2.1** 添加 JWT 认证

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The trigger list includes "make a plan" and "plan this out," which are common phrases users may say in many contexts and could cause unintended skill activation. Although examples and some non-examples are provided, the keyword trigger itself is not narrowly scoped to this specific skill or invocation context.
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 · README.md (reported line 109)May include surrounding context.
md
| Rule | If violated | |------|-------------| | Never skip approval (Phase 3) | Halt, undo actions, return to Phase 3 | | No research during planning | Discard results, plan structure only | | Must ask execution mode | Pause all work, wait for user reply | | Always update task status | Immediately fix plan doc |

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 94% confidence
Finding: The trigger set includes broad phrases like "make a plan" and auto-detection for complex requests, which can cause PAD mode to activate when the user did not intend to invoke this higher-agency workflow. In this skill, activation changes execution behavior, creates persistent plan files, and may later lead to tool use or sub-agent orchestration, so accidental invocation meaningfully expands the action surface.
Content: No source excerpt is available for this finding.

Session Persistence

Medium

Category: Rogue Agent
Confidence: 83% confidence
Finding: The skill persists detailed plan documents under a fixed workspace path and supports resuming prior plans, which creates durable session state that may contain sensitive user requests, task history, or operational context. Persistent cross-session artifacts increase the risk of unintended data retention, later disclosure, and context bleed between tasks if plans are resumed or searched broadly.
Content: Scanner excerpt · SKILL.md (reported line 27)May include surrounding context.
md
Violating the rules below is considered an execution failure. Treat the 🛑 STOP points as hard blockers, not suggestions. 1. **Skip Phase 3 Approve and jump to execution** → Halt immediately. Undo any actions taken. Return to Phase 3, update status to `🔵 Confirmed`, and wait for explicit user approval. 2. **Do deep research during Plan phase** → Discard all research results. The Plan phase is for structure only. No tool calls, no web searches, no file reads beyond the template and the minimum files required to create or resume the plan. 3. **Start executing without asking execution mode** → Pause all execution. Ask the user Foreground/Background, wait for reply, then resume. 4. **Complete tasks without updating plan file status** → Immediately update the plan doc. Each task must reflect its actual state (`🔄 In Progress` → `✅ Done` / `❌ Failed`). 5. **Mark a task done without verification** → Reopen the task immediately. In Phase 4, every task must follow `Execute → Verify → Mark done`.

Autonomous Decision Making

Medium

Category: Excessive Agency
Confidence: 75% 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 28)May include surrounding context.
md
1. **Skip Phase 3 Approve and jump to execution** → Halt immediately. Undo any actions taken. Return to Phase 3, update status to `🔵 Confirmed`, and wait for explicit user approval. 2. **Do deep research during Plan phase** → Discard all research results. The Plan phase is for structure only. No tool calls, no web searches, no file reads beyond the template and the minimum files required to create or resume the plan. 3. **Start executing without asking execution mode** → Pause all execution. Ask the user Foreground/Background, wait for reply, then resume. 4. **Complete tasks without updating plan file status** → Immediately update the plan doc. Each task must reflect its actual state (`🔄 In Progress` → `✅ Done` / `❌ Failed`). 5. **Mark a task done without verification** → Reopen the task immediately. In Phase 4, every task must follow `Execute → Verify → Mark done`. 6. **Deliver without button/text confirmation** → Do NOT auto-archive. Send the completion summary with buttons (or text fallback) and wait for user response.

Autonomous Decision Making

Medium

Category: Excessive Agency
Confidence: 75% 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 30)May include surrounding context.
md
2. **Do deep research during Plan phase** → Discard all research results. The Plan phase is for structure only. No tool calls, no web searches, no file reads beyond the template and the minimum files required to create or resume the plan. 3. **Start executing without asking execution mode** → Pause all execution. Ask the user Foreground/Background, wait for reply, then resume. 4. **Complete tasks without updating plan file status** → Immediately update the plan doc. Each task must reflect its actual state (`🔄 In Progress` → `✅ Done` / `❌ Failed`). 5. **Mark a task done without verification** → Reopen the task immediately. In Phase 4, every task must follow `Execute → Verify → Mark done`. 6. **Deliver without button/text confirmation** → Do NOT auto-archive. Send the completion summary with buttons (or text fallback) and wait for user response. 7. **Use oversized tasks when the work can be reasonably split** → Stop and decompose the task before continuing. PAD tasks should be small enough to track, verify, and recover from failure cleanly. 8. **Skip required review for high-risk work** → Pause before completion and trigger review or explicit self-review.

Autonomous Decision Making

Medium

Category: Excessive Agency
Confidence: 70% 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 184)May include surrounding context.
md
🛑 **STOP.** Do NOT auto-archive. Send confirmation and WAIT for user response: - If channel supports buttons: send `✅ Archive` / `🔧 Changes Needed` buttons, wait for click - If text only: send summary, wait for user reply matching "archive"/"done"/"looks good" or "changes"/"modify"/"needs work" - If no response: do nothing. Do NOT assume approval from silence. 4. If user clicks **Archive** (or confirms via text): - Update status to `✅ Completed`

Natural-Language Policy Violations

Low

Category: Not specified by scanner
Confidence: 89% confidence
Finding: The approval text examples include the Chinese term "确认" alongside English examples, but the document does not state that users may respond in their preferred language or that multilingual matching is optional. This can be read as prescribing accepted approval language forms rather than offering a language-neutral or user-choice approach.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.