Back to skill

Security audit

IBCAT - International Bilingual Chinese Advanced Translator

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious with high confidence because scanners report persistent changes that weaken checks in an installed dependency, broad process cleanup, and potentially destructive workspace cleanup.

Consider the skill’s persistent dependency changes, host Python installation instructions, and cleanup actions before installing. The Skillspector scan is marked partial: it inspected all listed components but could not resolve several referenced paths, so the evidence is incomplete. AIG completed with no reported results, which does not establish that the skill is safe.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (21)

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
85% 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 125)May include surrounding context.

md
## ⚠️ 注意事项 / Important Notes

- **必须清除翻译缓存 / Must clear translation cache:** `rm -f /root/.cache/babeldoc/cache.v1.db`
- **字体补丁需在 BabelDOC 更新后重新应用 / Font patches must be reapplied after BabelDOC updates**
- **需要系统安装 Noto CJK 字体 / Requires Noto CJK fonts installed on the system**

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The workflow explicitly directs persistent changes to an installed dependency that disable font and CMap integrity checks. Although documented rather than concealed, this weakens verification for subsequent executions and can allow substituted assets to be accepted if an attacker can modify the cache.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
92% confidence
Finding

The code supports the declared workflow but implements only its extraction stage, not the advertised translation and final bilingual PDF generation. This is a functional scope mismatch, not evidence of malicious behavior. Dependency installation, cache cleanup, font setup, and local proxy management are supporting implementation details rather than independent mismatches. Other skill components may implement the missing functionality, but that cannot be established from this chunk.

Content

No source excerpt is available for this finding.

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
86% confidence
Finding

The command unconditionally deletes extracted segments and completed translations from a fixed shared workspace. Although this is recognizable cleanup rather than malicious tool abuse, another run's work can be lost because the workflow does not establish exclusive ownership of those files.

Content

Scanner excerpt · SKILL.md (reported line 138)May include surrounding context.

bash
# 清理旧数据 / Clean old data
rm -f /data/user/work/segments.json /data/user/work/translations.json /data/user/work/unknown_segments.json

# 启动代理服务器(extract 模式)/ Start proxy server (extract mode)
PROXY_MODE=extract python3 scripts/proxy_server.py &

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
95% 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 · SKILL.md (reported line 131)May include surrounding context.

清除旧缓存(重要!避免使用过期翻译缓存)

Clear old cache (IMPORTANT! Avoid using stale translation cache)

rm -f /root/.cache/babeldoc/cache.v1.db

text

#### Step 2: 启动翻译代理(提取模式)/ Start Translation Proxy (Extract Mode)

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
95% 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 · SKILL.md (reported line 212)May include surrounding context.

md
sleep 2

# 清除 BabelDOC 翻译缓存(重要!)/ Clear BabelDOC translation cache (CRITICAL!)
rm -f /root/.cache/babeldoc/cache.v1.db

# 运行 BabelDOC 生成最终 PDF / Run BabelDOC to generate final PDF
babeldoc \

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
95% 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 · SKILL.md (reported line 257)May include surrounding context.

English: Before each re-translation, you MUST clear BabelDOC's translation cache database, otherwise stale translations from the previous run will be used:

bash
rm -f /root/.cache/babeldoc/cache.v1.db

字体补丁持久性 / Font Patch Persistence

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
95% 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 · scripts/render_pdf.sh (reported line 54)May include surrounding context.

sh
# Step 1: Clear BabelDOC translation cache (CRITICAL!)
echo "[Step 1] Clearing BabelDOC translation cache..."
rm -f /root/.cache/babeldoc/cache.v1.db
echo "  Cache cleared."

# Step 2: Kill any existing proxy

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
95% 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 · scripts/translate_pdf.sh (reported line 84)May include surrounding context.

sh
# Step 2: Clear old caches
echo "[Step 2] Clearing old caches..."
rm -f /root/.cache/babeldoc/cache.v1.db
mkdir -p "$WORK_DIR"
rm -f "$WORK_DIR/segments.json" "$WORK_DIR/translations.json" "$WORK_DIR/unknown_segments.json"

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
95% confidence
Finding

Every run unconditionally deletes existing segment and translation files in the selected working directory, including potentially valuable translations produced by earlier LLM work. The shared default directory and absence of confirmation or backup make accidental cross-run data loss plausible; the issue is unsafe destructive workspace handling, not shell injection, because the paths are quoted.

Content

Scanner excerpt · scripts/translate_pdf.sh (reported line 86)May include surrounding context.

sh
echo "[Step 2] Clearing old caches..."
rm -f /root/.cache/babeldoc/cache.v1.db
mkdir -p "$WORK_DIR"
rm -f "$WORK_DIR/segments.json" "$WORK_DIR/translations.json" "$WORK_DIR/unknown_segments.json"

# Step 3: Kill any existing proxy on the port
kill $(lsof -ti:$PROXY_PORT 2>/dev/null) 2>/dev/null || true

The translation workflow lacks a privacy warning about sending PDF content to an LLM provider.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The skill's workflow directs extracted PDF segments to an LLM for translation, and the README advertises support for arbitrary LLMs through a compatible proxy. It does not warn that choosing a remotely hosted provider sends document contents outside the local environment, potentially exposing confidential or personal information.

Content

No source excerpt is available for this finding.

The installation command bypasses system Python package protections without warning about the consequences.

Medium
Category
Output Handling
Confidence
98% confidence
Finding

The installation command explicitly bypasses externally managed Python environment protections. Running it against the host interpreter can replace distribution-managed dependencies and disrupt other applications, although this is not evidence of malicious intent.

Content

No source excerpt is available for this finding.

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

Dependency installation bypasses system-package protections without warning about effects on the host Python environment.

Medium
Category
Output Handling
Confidence
95% confidence
Finding

The dependency installation commands bypass protections intended to prevent pip from modifying a distribution-managed Python environment. Following these instructions on the host can introduce incompatible dependencies and affect applications outside the PDF workflow.

Content

No source excerpt is available for this finding.

Proxy cleanup can terminate unrelated processes without warning.

Medium
Category
Output Handling
Confidence
98% confidence
Finding

The cleanup command selects processes solely through their association with port 8899, without verifying ownership by this workflow. It can terminate an unrelated service or client process that the executing user has permission to signal.

Content

No source excerpt is available for this finding.

Tainted flow: 'tmp' from os.environ.get (line 59, credential/environment) → open (file write)

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/proxy_server.py (reported line 60)May include surrounding context.

python
def save_json(filepath, data):
    """Save JSON data atomically."""
    tmp = filepath + '.tmp'
    with open(tmp, 'w') as f:
        json.dump(data, f, ensure_ascii=False, indent=2)
    os.replace(tmp, filepath)

The renderer terminates processes using the selected port without verifying that they belong to the translation skill.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The script signals every process returned by lsof for the selected port without establishing that it is a translation proxy. A port collision or an inappropriate --port value can therefore terminate unrelated processes accessible to the executing user.

Content

No source excerpt is available for this finding.

The render step recursively deletes the existing working directory without disclosing that cleanup.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The renderer recursively deletes an existing babeldoc_render directory beneath the user-selected work directory without verifying that its contents belong to this invocation. Quoting prevents shell expansion but does not prevent loss of preexisting files or interference with another render using the same workspace.

Content

No source excerpt is available for this finding.

The font setup script also disables CMap integrity verification in the installed BabelDOC dependency.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The patch replaces CMap hash verification with an existence check, causing BabelDOC to accept corrupted or substituted cached CMaps. Exploitation requires an attacker-controlled cache entry, but the persistent dependency modification removes a safeguard for subsequent translation runs and can compromise document text mapping or disrupt processing. Offline font substitution does not require disabling integrity checks for unrelated CMap assets.

Content

No source excerpt is available for this finding.

Terminates existing processes based solely on their use of the configured proxy port.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The script terminates processes identified by port usage without checking that they belong to this translation workflow. A port collision or supplied --port value can therefore stop unrelated services accessible to the executing user, with broader disruption possible when run as root.

Content

No source excerpt is available for this finding.

Documentation claims render-phase automation, but the script implements only extraction.

Low
Category
Excessive Agency
Confidence
100% confidence
Finding

The header explicitly says the script automates both extraction and rendering. However, the argument parser at L29-L48 rejects the --render-only option recommended at L150-L160, and the script exits after extraction without implementing a render phase.

Content

No source excerpt is available for this finding.