Back to skill

Security audit

pikpak-batch-backup

Security checks for vulnerabilities and agentic risk

Overview

This skill copies PikPak files locally with clear safeguards, although its description advertises backup features it does not provide.

Use a trusted PikPak CLI and local environment, specify the intended folder explicitly, and review a dry run before approving downloads. Choose a fresh destination and protect the downloaded files and manifest. Resume, manifest input, and --json output are unavailable; local hashes do not prove equality with remote content. The inspected artifacts do not substantiate the malicious pre-scan signal. Downloads consume [PikPak connected-app quota](https://mypikpak.com/zh-CN/help-center/connected_apps/webdav/speed_limit).

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
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (9)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The discovery description promises resumable and manifest-driven backups, while the detailed documentation explicitly disclaims resumption and describes no manifest-input workflow. This can cause users or agents selecting the skill from metadata to rely on unsupported backup capabilities, although documented rejection of existing files and nonzero failure reporting reduce the risk of silent loss; no destructive or malicious behavior is established.

Content

No source excerpt is available for this finding.

Direct flow: os.environ.get (credential/environment) → subprocess.run (code execution)

High
Category
Data Flow
Confidence
80% confidence
Finding

Data flows directly from a source (env vars, files, network) to a sink (network output, exec, file write) without intermediate validation.

Content

Scanner excerpt · scripts/batch_backup.py (reported line 106)May include surrounding context.

python
def pikpak(*args: str, timeout: int = 3600):
    try:
        proc = subprocess.run([os.environ.get("PIKPAK_BIN", "pikpak"), *args],
                              capture_output=True, text=True, timeout=timeout)
        if proc.returncode:
            return proc.returncode, "", f"cli_failed (exit_code={proc.returncode})"

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env, file_write, shell.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

The manifest advertises resumable backups, but the documented implementation explicitly does not support resuming.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The description explicitly advertises resume support, but the body states that there is no incremental or interrupted-transfer recovery and that reruns require a new directory. This is the same misleading-capability issue captured by TP4, not an additional exploit; its main risk is misplaced reliance and wasted transfer, storage, or quota after interruptions.

Content

No source excerpt is available for this finding.

The manifest advertises resume support, but the implementation documentation explicitly states that resume is not implemented.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The skill description explicitly advertises resume support in English and Chinese, while the documentation states that byte-level resume and checkpoint recovery are not implemented. This capability mismatch can cause users or agents to plan backups around recovery guarantees that do not exist, resulting in repeated downloads or incomplete backup coverage after interruption. The documented non-overwrite behavior, failure reporting, and explicit limitations reduce the danger; this is a misleading capability claim, not evidence of arbitrary code execution or silent data corruption.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/batch_backup.py (reported line 106)May include surrounding context.

python
def pikpak(*args: str, timeout: int = 3600):
    try:
        proc = subprocess.run([os.environ.get("PIKPAK_BIN", "pikpak"), *args],
                              capture_output=True, text=True, timeout=timeout)
        if proc.returncode:
            return proc.returncode, "", f"cli_failed (exit_code={proc.returncode})"

Tainted flow: 'timeout' from os.environ.get (line 158, credential/environment) → subprocess.run (code execution)

Medium
Category
Data Flow
Confidence
65% confidence
Finding

Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.

Content

Scanner excerpt · scripts/batch_backup.py (reported line 106)May include surrounding context.

python
def pikpak(*args: str, timeout: int = 3600):
    try:
        proc = subprocess.run([os.environ.get("PIKPAK_BIN", "pikpak"), *args],
                              capture_output=True, text=True, timeout=timeout)
        if proc.returncode:
            return proc.returncode, "", f"cli_failed (exit_code={proc.returncode})"

The manifest promises resume support, but existing backup files cause an abort and interrupted transfers cannot be resumed.

Medium
Category
Excessive Agency
Confidence
100% confidence
Finding

The skill description explicitly advertises resume support (断点续传), but L252–L257 reject existing destination files or a backup record instead of using them to resume. L363 explicitly confirms that resume is unsupported, so an interrupted backup cannot be continued as advertised.

Content

No source excerpt is available for this finding.

The manifest advertises manifest-driven backups, but the command only supports folder-driven selection.

Low
Category
Excessive Agency
Confidence
98% confidence
Finding

The description promises backups 'by folder or manifest,' but the argument parser provides no backup-manifest input. The only file-input option is a listing fixture restricted to dry runs at L275–L276; the backup manifest written later is an output record, not an executable backup input.

Content

No source excerpt is available for this finding.