Back to skill

Security audit

角色称呼与别名核对

Security checks for vulnerabilities and agentic risk

Overview

This skill provides a narrowly scoped manuscript consistency check, with no evidence of harmful actions or hidden access.

Review the installer and source revision before installation. Treat results as suggestions: unapproved address forms can be labeled incorrect, while uniquely registered narrative names may pass without a target ID. Confirm each proposed change yourself.

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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (15)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The only supplied command performs skill installation, not the declared manuscript review. It contains no logic for character attribution, relationship validation, ambiguity detection, or replacement suggestions. Although this may be an installation instruction for a relevant skill, the installed skill's implementation is not provided, so its behavior cannot be verified from this snippet.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
88% confidence
Finding

The installation command uses an unpinned npm executable, so running it may download and execute a changing version of the skills package with the user's privileges. A compromised release could execute malicious code; this is an installation supply-chain risk, not evidence of an MCP server or malicious skill behavior.

Content

No source excerpt is available for this finding.

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.en.md (reported line 21)May include surrounding context.

md
## Source and license

The use case was reviewed against [story-skills character-management](https://github.com/danjdewhurst/story-skills/blob/main/skills/character-management/SKILL.md). This repository's wording, examples, and implementation are original; no external text or source was copied. See [LICENSE](LICENSE).

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.md (reported line 31)May include surrounding context.

md
## 来源与许可

业务用途参考了 [story-skills 的 character-management](https://github.com/danjdewhurst/story-skills/blob/main/skills/character-management/SKILL.md),本仓库说明、示例和实现均为原创,未复制其正文或源码。详见 [LICENSE](LICENSE)。

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · sources/story-skills.md (reported line 3)May include surrounding context.

md
# 采用来源记录

- URL: https://github.com/danjdewhurst/story-skills/blob/main/skills/character-management/SKILL.md
- 核对日期:2026-10-04
- 实际采用:仅将“角色管理”作为业务领域参考,未复制任何文本、代码或资源;本 Skill 的输入格式、检查边界、脚本与示例均为原创。
- 许可证:未引入其受版权保护的内容;本仓库自身以 MIT 发布。

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
91% confidence
Finding

The installation command runs an unpinned npm CLI through npx and does not specify an immutable revision of the skill repository. Although this is not an MCP server reference as the detector claims, changing or compromised upstream releases could execute unwanted code with the installing user's privileges or introduce malicious skill instructions; the README provides no evidence of an actual compromise.

Content

No source excerpt is available for this finding.

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

Unregistered address forms are classified as wrong rather than pending confirmation when ownership is unknown.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

When an address has no registered owner and no matching approval rule, the code classifies it as wrong even though the skill promises pending confirmation for missing alias ownership. This conflates incomplete identity information with a confirmed relationship violation and can mislead editorial decisions, although the recommendation to preserve the text pending confirmation limits the impact.

Content

No source excerpt is available for this finding.

A uniquely registered name without a target ID is silently accepted instead of marked pending confirmation.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

For a non-address mention with one registered owner but no target ID, none of the final branches emits a finding, silently treating the mention as acceptable. This violates the stated requirement to mark missing targets as pending confirmation and can conceal a reference to the wrong character, but the impact is limited to editorial report integrity rather than code execution or data exposure.

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/self_test.py (reported line 7)May include surrounding context.

python
ROOT = Path(__file__).resolve().parents[1]

def run(name):
    out = subprocess.check_output([sys.executable, str(ROOT/'scripts/check_names.py'), str(ROOT/'fixtures'/name)], text=True)
    return json.loads(out)

def main():

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/self_test.py (reported line 35)May include surrounding context.

python
}
    with tempfile.NamedTemporaryFile('w', encoding='utf-8', suffix='.json') as f:
        json.dump(sample, f, ensure_ascii=False); f.flush()
        out = json.loads(subprocess.check_output([sys.executable, str(ROOT/'scripts/check_names.py'), f.name], text=True))
    assert not any(row[0] in {'x1','x2','x3'} for row in out['findings']), out['findings']
    assert {row[0] for row in out['findings']} == {'x4','x5','x6','x7','x8'}, out['findings']
    assert all(row[2] == '待确认' for row in out['findings'] if row[0] in {'x4','x5'}), out['findings']

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/self_test.py (reported line 53)May include surrounding context.

python
duplicate_address['mentions'] = [{'anchor':'x9','token':'同名','kind':'address','speaker':'s','target':'c'}]
    with tempfile.NamedTemporaryFile('w', encoding='utf-8', suffix='.json') as f:
        json.dump(duplicate_address, f, ensure_ascii=False); f.flush()
        duplicate_address_out = json.loads(subprocess.check_output([sys.executable, str(ROOT/'scripts/check_names.py'), f.name], text=True))
    duplicate_address_finding = duplicate_address_out['findings'][0]
    assert duplicate_address_finding[2] == '错称', duplicate_address_out['findings']
    assert '甲、乙' in duplicate_address_finding[3] and '丙' in duplicate_address_finding[3], duplicate_address_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/self_test.py (reported line 45)May include surrounding context.

python
duplicate_mismatch['mentions'] = [{'anchor':'x8','token':'同名','kind':'name','target':'c'}]
    with tempfile.NamedTemporaryFile('w', encoding='utf-8', suffix='.json') as f:
        json.dump(duplicate_mismatch, f, ensure_ascii=False); f.flush()
        mismatch_out=json.loads(subprocess.check_output([sys.executable,str(ROOT/'scripts/check_names.py'),f.name],text=True))
    assert mismatch_out['findings'][0][2] == '错称' and '全局替换' in mismatch_out['findings'][0][4], mismatch_out['findings']
    # Multiple owners do not make an address pending when its explicit target
    # is a third person: none of those owners can be the target.

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/self_test.py (reported line 69)May include surrounding context.

python
missing_speaker=json.loads(json.dumps(sample)); missing_speaker['mentions']=[{'anchor':'x7','token':'宝贝','kind':'address','target':'a'}]
    with tempfile.NamedTemporaryFile('w', encoding='utf-8', suffix='.json') as f:
        json.dump(missing_speaker, f, ensure_ascii=False); f.flush()
        missing_out=json.loads(subprocess.check_output([sys.executable,str(ROOT/'scripts/check_names.py'),f.name],text=True))
    assert missing_out['findings'][0][3] == '缺少说话人 ID'
    print('PASS: approved triples pass; unresolved shared names/aliases and unknown IDs await confirmation.')
if __name__ == '__main__': main()

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

Scanner excerpt · LICENSE (reported line 12)May include surrounding context.

text
permit persons to whom the Software is furnished to do so.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A
PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE