Back to skill

Security audit

前期会计差错更正·逐条挂源

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marked this skill suspicious with medium confidence, mainly because scanners report external document parsing without a privacy warning and unrelated remote service discovery and local usage tracking.

Consider the suspicious verdict and incomplete Skillspector evidence before installing. Review whether sending documents to an external parser is acceptable for your material, and whether you are comfortable with remote tool discovery and a persistent local usage marker. The AIG scan completed without findings, but Skillspector marked its analysis partial because a referenced local path was missing, so the results do 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
Findings (25)

YARA rule 'agent_skill_destructive_autonomous_actions': Autonomous destructive filesystem, shell history, or repository actions in AI agent skills [agent_skills]

High
Category
YARA Match
Confidence
75% confidence
Finding

YARA rule matched a known malware signature (reverse shell, backdoor, ransomware, C2 framework, or info stealer).

Content

Scanner excerpt · README.en.md (reported line 147)May include surrounding context.

md
**empty file** at `~/.cue-usage/<slug>-<date>`.

**Why:** both platform download counters are unreliable — a new listing is seeded with 50–60 initial counts in its first 48 hours, and those are not users. This local file is the only evidence that someone actually installed and ran the piece.

**It never gets in the way:** any failure (read-only home directory, full quota, no HOME) is skipped silently — no error, no change to the exit code, not a word of noise. Running several times in one day leaves one file.

To clear it: `rm -rf ~/.cue-usage`.

---

## 9. Files

| File | Purpose |
|---|---|
| `SKILL.md` / `SKILL.en.md` | Chinese and English skill descriptions and checker contract |
| `scripts/check_correction.py` | five checks; `--selftest` runs fixtures, `--demo` prints a sample |
| `scripts/fixtures/` | one good sample, one bad sample, three-tier runner |
| `tests/test_correction.py` | package tests (entry contract, zero input, beacon) |

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
90% 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.en.md (reported line 149)May include surrounding context.

md
**It never gets in the way:** any failure (read-only home directory, full quota, no HOME) is skipped silently — no error, no change to the exit code, not a word of noise. Running several times in one day leaves one file.

To clear it: `rm -rf ~/.cue-usage`.

---

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
90% 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 150)May include surrounding context.

md
**它不会碍事:** 任何失败(家目录只读、配额满、无 HOME)都静默跳过,不报错、不改退码、不吐一句杂音。同一天跑多次只留一枚。

要清掉:`rm -rf ~/.cue-usage`。

---

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
90% 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.en.md (reported line 149)May include surrounding context.

md
**It never gets in the way:** any failure (read-only home directory, full quota, no HOME) is skipped silently — no error, no change to the exit code, not a word of noise. Running several times in one day leaves one file.

To clear it: `rm -rf ~/.cue-usage`.

---

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.en.md (reported line 149)May include surrounding context.

md
**It never gets in the way:** any failure (read-only home directory, full quota, no HOME) is skipped silently — no error, no change to the exit code, not a word of noise. Running several times in one day leaves one file.

To clear it: `rm -rf ~/.cue-usage`.

---

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 150)May include surrounding context.

md
**它不会碍事:** 任何失败(家目录只读、配额满、无 HOME)都静默跳过,不报错、不改退码、不吐一句杂音。同一天跑多次只留一枚。

要清掉:`rm -rf ~/.cue-usage`。

---

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied shell commands perform network-based service and tool discovery, not extraction or verification of accounting-error corrections. The chunk does not process financial documents, produce source-linked entries, or handle restatement amounts. Nothing supplied establishes these network requests as supporting implementation details of the declared workflow. No document exfiltration or remote tool execution beyond discovery is shown.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The document-validation functionality supports the declared accounting workflow and is not itself a material mismatch. However, mark_first_run() records daily execution evidence in the user's home directory, expressly for usage measurement. This persistent side effect is unrelated to extracting or checking accounting disclosures and is not disclosed in the declared purpose or permissions. The supplied code shows no network transmission; the mismatch is local usage tracking, not exfiltration.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

CLI, fixture, and output-validation tests are supporting implementation details consistent with the declared accounting task. The usage beacon is different: recording when the skill runs is not part of extracting or validating accounting corrections, and its home-directory tracking behavior is not disclosed in the description. The checker implementation is not supplied, so the beacon is evidenced by explicit test expectations rather than its write implementation. No network transmission or credential access is shown.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The supplied chunk performs remote service and tool discovery rather than analyzing financial disclosures or producing a source-linked correction register. Tool discovery could support a larger implementation, but this chunk provides no accounting-specific connection that would establish it as a supporting detail of the declared workflow. There is no evidence here of document upload, financial-data exfiltration, or execution of the discovered tools.

Content

No source excerpt is available for this finding.

Env Variable Harvesting

High
Category
Data Exfiltration
Confidence
60% confidence
Finding

Code enumerates, copies, or searches environment variables for secrets. Bulk environment access can collect credentials unrelated to the skill's stated purpose.

Content

Scanner excerpt · tests/test_correction.py (reported line 31)May include surrounding context.

python
def run(*args, env=None, cwd=None):
    e = dict(os.environ, PYTHONDONTWRITEBYTECODE="1")
    if env:
        e.update(env)
    return subprocess.run([sys.executable, "-B", str(args[0]), *map(str, args[1:])],

Env Variable Harvesting

High
Category
Data Exfiltration
Confidence
60% confidence
Finding

Code enumerates, copies, or searches environment variables for secrets. Bulk environment access can collect credentials unrelated to the skill's stated purpose.

Content

Scanner excerpt · tests/test_correction.py (reported line 134)May include surrounding context.

python
os.chmod(td, 0o700)

    def test_no_home_is_silent(self):
        e = dict(os.environ, PYTHONDONTWRITEBYTECODE="1")
        e.pop("HOME", None)
        r = subprocess.run([sys.executable, "-B", str(CHECK), "--demo"],
                           capture_output=True, text=True, env=e, cwd=str(PKG))

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · README.en.md (reported line 52)May include surrounding context.

bash
CATALOG="https://cuecue.cn/api/mcp-catalog"
curl -s "$CATALOG" | head -c 2000
curl -s -X POST "$ROUTING_URL" \
  -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · README.md (reported line 53)May include surrounding context.

CATALOG="https://cuecue.cn/api/mcp-catalog" curl -s "$CATALOG" | head -c 2000

从目录里取本组的路由地址,确认工具在场

curl -s -X POST "$ROUTING_URL"
-H 'Content-Type: application/json'
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

text

The document-parsing workflow lacks a privacy warning about sending user-supplied documents to an external service.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The workflow directs user-provided documents through a parsing service without clearly warning that their contents may leave the local environment or requiring upload approval. Published filings reduce the risk, but private drafts or annotated financial documents could contain confidential information.

Content

No source excerpt is available for this finding.

The document-parsing workflow lacks a privacy warning about sending user-provided financial documents to an external service.

Medium
Category
Output Handling
Confidence
86% confidence
Finding

The user-supplied document route invokes a parsing channel without an explicit warning about external content processing or an upload-consent gate. The usual public-filing context limits exposure, but private drafts and annotations could be disclosed unintentionally.

Content

No source excerpt is available for this finding.

The attachment-parsing workflow lacks a warning that document content is sent to an external service.

Medium
Category
Output Handling
Confidence
93% confidence
Finding

The attachment workflow starts the external parsing leg based on document format without explicitly obtaining approval to transmit document content. Mentioning service billing and configuration does not explain the privacy boundary, particularly for unpublished or annotated financial material.

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.

The workflow sends attachments to an external parsing service without a privacy warning.

Medium
Category
Output Handling
Confidence
89% confidence
Finding

The workflow directs attachments and scanned documents to an external parsing service without requiring a confidentiality check or informed approval for that transfer. Published disclosures reduce the risk, but user-supplied screenshots or unpublished financial attachments could expose confidential information outside the user's environment.

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 · tests/test_correction.py (reported line 34)May include surrounding context.

python
e = dict(os.environ, PYTHONDONTWRITEBYTECODE="1")
    if env:
        e.update(env)
    return subprocess.run([sys.executable, "-B", str(args[0]), *map(str, args[1:])],
                          capture_output=True, text=True, env=e, cwd=str(cwd or PKG))

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 · tests/test_correction.py (reported line 136)May include surrounding context.

python
def test_no_home_is_silent(self):
        e = dict(os.environ, PYTHONDONTWRITEBYTECODE="1")
        e.pop("HOME", None)
        r = subprocess.run([sys.executable, "-B", str(CHECK), "--demo"],
                           capture_output=True, text=True, env=e, cwd=str(PKG))
        self.assertEqual(r.returncode, 0)
        self.assertEqual(r.stderr, "", f"无 HOME 不许吐杂音:{r.stderr[:200]}")

The documented first-run beacon writes usage-tracking files outside the accounting-analysis workflow.

Low
Category
Excessive Agency
Confidence
97% confidence
Finding

The documented beacon automatically persists the skill identifier and execution date outside the workspace without an explicit consent mechanism. This creates a small local usage-privacy exposure unrelated to accounting analysis, although no transmission or document-content collection is shown.

Content

No source excerpt is available for this finding.

L128–L142 implement persistent usage tracking unrelated to accounting-correction analysis.

Low
Category
Excessive Agency
Confidence
100% confidence
Finding

mark_first_run creates a persistent, dated usage marker in the user's home directory without an opt-in gate and suppresses all write failures. This exposes limited local execution metadata and adds an unrelated filesystem side effect, but the code shows no network reporting or document-content collection.

Content

No source excerpt is available for this finding.

L136 indirectly reads environment variables despite L003 explicitly claiming that none are read.

Low
Category
Excessive Agency
Confidence
99% confidence
Finding

The module documentation at L003 promises 不读环境变量 (no environment-variable reads). However, os.path.expanduser("~") at L136 consults home-directory environment variables, such as HOME on Unix, and this path is executed through the unconditional call at L657.

Content

No source excerpt is available for this finding.

L108–L114 explicitly require an undeclared usage marker in the user's home directory.

Low
Category
Excessive Agency
Confidence
94% confidence
Finding

These tests explicitly require the checker to persist a dated skill-usage marker under HOME/.cue-usage and to suppress marker failures, establishing a tracking-related side-effect contract unrelated to document validation. The shown marker tests use temporary homes and do not demonstrate network transmission or storage of document contents, limiting the supported impact to local usage privacy and unexpected persistence.

Content

No source excerpt is available for this finding.