Back to skill

Security audit

吞贼·净化魄

Security checks for vulnerabilities and agentic risk

Overview

The skill is a self-healing guide, but it tells agents to clean files, roll back state, trigger GC, and restart processes without enough limits or confirmation.

Review before installing. Use it only in constrained workspaces or non-production environments unless the agent must ask before deleting files, rolling back versions, triggering recovery actions, or restarting processes. Prefer read-only diagnosis by default, require exact target lists and backups for changes, and restrict any process or filesystem access to explicit allowlists.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • 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
Findings (1)

T05 · Unauthorized Access and Privilege Escalation

Error
Location
SKILL.md:32
Finding

Unrestricted Destructive System-Recovery Operations

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 32–36, 49–56, 101–105, 111–120
Vulnerability Type: Broadly authorized cleanup, rollback, garbage collection, and process-restart operations without least-privilege safeguards
Risk Level: High

Relevant Skill Directives

The following is an English rendering of the complete relevant directives:

markdown
### 3. Redundancy Cleanup
- Duplicate file detection
- Temporary file cleanup
- Resource release

When the system encounters an exception or the user requests a health check:

User: Clean up redundant files
→ Activate Auto Healer
→ Perform cleanup

### Rollback Strategy
Version records: Retain the most recent N versions
One-click rollback: Support rapid rollback
Data consistency: Ensure data remains complete after rollback

### Memory Leak Detection
- Monitor memory usage trends
- Automatically trigger garbage collection after exceeding a threshold
- Record leak patterns

### Process Exception Detection
- Heartbeat monitoring
- Restart unresponsive processes
- Resource exhaustion protection

Technical Analysis

The Skill authorizes an agent to perform state-changing system operations, including deleting duplicate or temporary files, releasing resources, rolling back versions, triggering garbage collection, and restarting processes. It does not define:

  • Permitted filesystem roots or file allowlists.
  • Criteria for determining whether a file is temporary or duplicated.
  • Protection against symbolic-link traversal or deletion outside an approved workspace.
  • Process ownership checks or an allowlist of restartable services.
  • Required user confirmation before each mutation.
  • Dry-run behavior or presentation of exact targets.
  • Backup, transaction, or restoration requirements before cleanup.
  • Privilege restrictions for filesystem and process-management tools.
  • Concrete thresholds and validation rules for automatic garbage collection.
  • Controls preventing rollback to an unsafe o ...[truncated 2070 chars]
Remediation
View remediation

Remediation Suggestions

  1. Make diagnosis read-only by default and separate diagnosis from remediation.
  2. Require explicit user approval immediately before each deletion, rollback, garbage-collection action, resource release, or process restart.
  3. Present an exact target list and anticipated impact before requesting approval.
  4. Restrict cleanup to configured, canonicalized workspace directories and explicitly approved temporary directories.
  5. Reject targets outside approved roots and prevent symbolic-link, hard-link, and path-traversal escapes.
  6. Define reliable file-classification criteria, retention periods, exclusions, and size or age thresholds.
  7. Use a dry-run mode by default and move files to a recoverable quarantine area instead of permanently deleting them.
  8. Create verified backups or snapshots before rollback and destructive cleanup.
  9. Permit process restarts only for an explicit allowlist, after validating process ownership, identity, dependencies, and restart limits.
  10. Run the Skill without elevated privileges and expose only the minimum tools required for the approved task.
  11. Require bounded resource thresholds, rate limits, cooldown periods, and circuit breakers for automatic recovery.
  12. Record targets, approvals, actions, outcomes, and restoration information in an audit log.
  13. Validate rollback compatibility and data integrity both before and after restoration.
  14. Stop and request operator intervention when target classification is ambiguous or when recovery affects shared or production resources.
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill promotes automatic cleanup and self-healing actions but does not warn that these operations may delete files, restart processes, alter configurations, or otherwise affect availability and data integrity. Users may interpret the capability as safe routine maintenance and authorize risky actions without informed consent.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The rollback and self-healing sections describe autonomous GC triggering, process restart, and rollback behavior without stating prerequisites, safeguards, or risks to data consistency and service uptime. In operational contexts, automated remediation that modifies live systems can amplify incidents if thresholds, targeting, or rollback criteria are wrong.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger language is broad enough that ordinary user statements about instability or health checks could invoke autonomous recovery behavior without clear confirmation boundaries. In a skill that advertises cleanup, retries, rollback, and self-healing, ambiguous activation increases the chance of unintended operational changes to system state.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

Nearly all user-facing descriptive content, examples, and operating guidance are written in Chinese, with no indication that users may choose another language. Under the policy, forcing a specific language without opt-in is a natural-language policy concern unless the locale restriction is clearly justified.

Content

No source excerpt is available for this finding.