Back to skill

Security audit

boc init

Security checks for vulnerabilities and agentic risk

Overview

This is a real deployment helper, but it asks for root SSH access and includes unsafe package verification, deletion, and password-tool instructions that could damage or compromise a deployment host.

Review this skill carefully before installing. Use a test or dedicated deployment host, avoid root password authentication where possible, prefer approved SSH/key or secret-manager workflows, verify deployment packages with trusted checksums or signatures, replace the sshpass download/install guidance with trusted managed tooling, and require confirmation or backups before deleting any existing deployment directory.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (4)

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:37
Finding

Deployment archive integrity verification is incorrectly implemented

Content
View full analysis
&& sha256sum .sha256 > /tmp/sha256sum-.txt ``` Periodically check the verification result: ```bash cat /tmp/sha256sum-.txt ``` Success criterion: the result contains ": Success" ``` ### Technical Analysis The command calculates the SHA-256 digest of the checksum file itself rather than validating the deployment archive against that checksum file. Proper GNU `sha256sum` verification ordinarily requires `sha256sum -c `. In addition, `nohup cd ` is not a reliable way to run this operation because `cd` is normally a shell builtin rather than a standalone executable that `nohup` can invoke directly. The command therefore may fail before verification occurs. The documented success criterion does not correspond to output that this command would normally produce. As a result, the workflow provides no reliable integrity guarantee before extracting the archive and executing its bundled `bocctl` program. ### Attack Path 1. An attacker with write access to the deployment location replaces or modifies the deployment archive. 2. The attacker may also replace the adjacent checksum file. 3. The Skill runs the documented command, which hashes the checksum file instead of validating the archive. 4. The workflow treats the ineffective or incorrectly interpreted result as package validation. 5. The modified archive is extracted. 6. The attacker-controlled `bocctl` executable is run through `./bocctl init`. ### Impact Assessment Successful exploitation can result in arbitrary code execution under the account used to initialize th ...[truncated 307 chars]
Remediation
View remediation

T03 · Remote Payload Retrieval and Execution

Error
Location
SKILL.md:148
Finding

Unverified password-handling executable is retrieved from a personal repository

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
SKILL.md:149
Finding

Unpinned and incorrectly specified Python dependency installation

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:37
Finding

User-controlled paths and filenames are interpolated into shell commands without validation

Content
View full analysis
&& sha256sum .sha256 > /tmp/sha256sum-.txt cat /tmp/sha256sum-.txt cd test -d BOC_k8s_noarch && rm -rf BOC_k8s_noarch nohup tar -xzf > /tmp/unpack.log 2>&1 & ps aux |grep 'tar -xzf ' ``` ### Technical Analysis The Skill declares `deploy_pkg_dir` and `deploy_pkg_file` as input parameters and substitutes them directly into shell command text. The placeholders are not quoted, canonicalized, constrained to safe characters, or passed as discrete process arguments. A value containing shell metacharacters such as semicolons, command substitutions, redirection operators, or line breaks can alter the command structure and execute additional commands. Values beginning with a hyphen can also be interpreted as options by utilities unless option parsing is terminated with `--`. The workflow additionally performs recursive deletion after changing directories. Although the deleted name is fixed, failure to strictly validate and confirm the canonical working directory can cause deletion in an unintended location. This exceeds a safe minimum-privilege implementation because the same operation can be performed with validated paths and structured process invocation. ### Attack Path 1. An attacker influences the package directory or filename supplied to the Skill, such as through a copied deployment request or untrusted automation input. 2. The attacker includes shell syntax in the parameter value. 3. The agent substitutes that value into one of the documented commands. 4. The remote shell parses the injected syntax as an additional command. 5. The injected command executes with the SSH account’s privileges. 6 ...[truncated 858 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Rogue AgentSelf-Modification, Session Persistence
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (9)

Chaining Abuse

High
Category
Tool Misuse
Confidence
96% confidence
Finding

The chained command test -d BOC_k8s_noarch && rm -rf BOC_k8s_noarch performs unconditional recursive deletion once the directory exists, without interactive confirmation or stronger path safety checks. In a remote root-operated deployment workflow, mistakes in working directory, variable substitution, or operator assumptions can turn this into destructive deletion of important filesystem contents.

Content

Scanner excerpt · SKILL.md (reported line 54)May include surrounding context.

bash
# 解压前先清理旧目录
cd <deploy_pkg_dir>
test -d BOC_k8s_noarch && rm -rf BOC_k8s_noarch

# 后台执行解压(30GB文件预计15-20分钟)
nohup tar -xzf <deploy_pkg_file> > /tmp/unpack.log 2>&1 &

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The README explicitly requires an SSH password as an input but provides no guidance on secure handling, storage, masking, or transmission of that credential. In a deployment-initialization skill that automates remote access, this increases the risk of operators exposing credentials in logs, chat transcripts, shell history, or insecure parameter passing.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
83% confidence
Finding

The README describes automated package verification, archive extraction, and execution of bocctl init on a remote deployment host without warning that these steps can make significant system changes. In infrastructure automation, omission of change-impact warnings can lead users to run the skill on the wrong host or with untrusted packages, causing service disruption or unsafe system modification.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill requests SSH credentials, including a plaintext password, and performs remote administrative actions without clearly warning about credential sensitivity or the impact on the target host. This increases the risk of mishandling secrets and executing invasive changes on production-like systems without informed consent.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
65% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 39)May include surrounding context.

执行 SHA256 校验:

bash
nohup cd <deploy_pkg_dir> && sha256sum <deploy_pkg_file>.sha256 > /tmp/sha256sum-<deploy_pkg_file>.txt

定期检查校验结果

bash

Session Persistence

Medium
Category
Rogue Agent
Confidence
65% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 57)May include surrounding context.

执行 SHA256 校验:

bash
nohup cd <deploy_pkg_dir> && sha256sum <deploy_pkg_file>.sha256 > /tmp/sha256sum-<deploy_pkg_file>.txt

定期检查校验结果

bash

Session Persistence

Medium
Category
Rogue Agent
Confidence
65% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 86)May include surrounding context.

执行 SHA256 校验:

bash
nohup cd <deploy_pkg_dir> && sha256sum <deploy_pkg_file>.sha256 > /tmp/sha256sum-<deploy_pkg_file>.txt

定期检查校验结果

bash

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill instructs deleting an existing directory with rm -rf BOC_k8s_noarch before extraction, but it does not clearly warn the user that existing contents will be permanently destroyed. In a deployment host context, that directory may contain prior installs, logs, or customized state, so silent cleanup can cause irreversible operational data loss.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
83% confidence
Finding

The natural-language instructions and examples are entirely in Chinese, which effectively imposes a language requirement on users. There is no indication that the skill offers multilingual support, asks the user for language preference, or is intentionally limited to a Chinese-only audience for a documented regional reason.

Content

No source excerpt is available for this finding.