Back to skill

Security audit

Accounting|记账 - 资产管理/信用卡/预购清单/年月统计

Security checks for vulnerabilities and agentic risk

Overview

The skill fits its local bookkeeping purpose, with disclosed file edits and some installation and reporting caveats.

Keep backups and raw bills. Default refund handling changes original amounts; request separate linked adjustment entries if you need transaction history. Use --overwrite only when intentionally replacing whole ledger files, and cross-check reimbursement-related tag breakdowns. Review and pin the installer and repository revision before following the README installation command.

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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • System Prompt LeakageDirect Leakage, Indirect Extraction, Tool-Based Exfiltration
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (11)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
93% confidence
Finding

The visible commands concern publication to named platforms rather than importing, recording, validating, or analyzing local financial data. This is materially different from the declared workflow. However, publish.py is not supplied, so the snippet does not establish what is published, whether network transfers occur, or whether any ledger files are uploaded.

Content

No source excerpt is available for this finding.

Direct Prompt Extraction

High
Category
System Prompt Leakage
Confidence
85% confidence
Finding

Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.

Content

Scanner excerpt · scripts/import_ledger.py (reported line 345)May include surrounding context.

python
def fallback_pair(config: SkillConfig, rules: ImportRules) -> tuple[str, str]:
    """The configured fallback, or the first category when the fallback is absent."""
    if rules.fallback and config.categories.accepts(*rules.fallback):
        return rules.fallback
    for category, subs in config.categories.map.items():
        for subcategory in sorted(subs):
            return category, subcategory

Direct Prompt Extraction

High
Category
System Prompt Leakage
Confidence
85% confidence
Finding

Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.

Content

Scanner excerpt · scripts/import_ledger.py (reported line 373)May include surrounding context.

python
if not config.categories.accepts(rule_category, rule_subcategory):
                continue
            if any(keyword.lower() in lowered for keyword in keywords):
                return rule_category, rule_subcategory, False
        if config.categories.accepts("退款", "退款"):
            return "退款", "退款", False
        return (*fallback_pair(config, rules), False)

Direct Prompt Extraction

High
Category
System Prompt Leakage
Confidence
85% confidence
Finding

Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.

Content

Scanner excerpt · scripts/import_ledger.py (reported line 388)May include surrounding context.

python
if tx_type == "收入":
        for rule_category, rule_subcategory, keywords in rules.category_keywords:
            if rule_category == "收入" and any(k.lower() in lowered for k in keywords):
                return rule_category, rule_subcategory, False
        if config.categories.accepts("收入", "其他"):
            return "收入", "其他", raw_category != "收入"
        return (*fallback_pair(config, rules), raw_category != "收入")

Direct Prompt Extraction

High
Category
System Prompt Leakage
Confidence
85% confidence
Finding

Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.

Content

Scanner excerpt · scripts/import_ledger.py (reported line 406)May include surrounding context.

python
if rule_category == "收入" or not config.categories.accepts(rule_category, rule_subcategory):
            continue
        if any(k.lower() in lowered for k in keywords):
            return rule_category, rule_subcategory, False
    return (*fallback_pair(config, rules), True)

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

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
89% confidence
Finding

The documented installation command runs an unpinned npm CLI (npx skills) and selects the skill repository without an immutable revision, exposing installation to upstream supply-chain changes. If either source is compromised, a user following this command could execute malicious code with their local permissions, potentially exposing sensitive financial files. The reference is an installer rather than an MCP server, so the static finding's label is inaccurate, but its underlying version-pinning concern is valid.

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 npx skills without a pinned package version, allowing a fresh installation to execute a changing third-party release with the user's privileges. Although this is an installer rather than an MCP server, the underlying supply-chain risk is real: a compromised release could access local financial files, and the skill's local-only privacy statements do not protect against malicious installer code. Nothing in the supplied README indicates malicious intent or an existing compromise.

Content

No source excerpt is available for this finding.

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

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.

Default refund and reimbursement handling overwrites original transaction amounts without warning about loss of historical data.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The documented default replaces original transaction amounts with net amounts, potentially discarding the original value and obscuring when refunds or reimbursements occurred. In a financial ledger, this can undermine auditability, historical cash-flow reconstruction, and gross-spending reports unless an independent history is retained. The document offers separate entries when original amounts must be preserved, indicating benign simplification rather than malicious intent, but does not require preservation before overwriting.

Content

No source excerpt is available for this finding.

The documented net-total formula contradicts the implemented reimbursement deduction.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The module docstring at L004–L005 explicitly defines net as gross minus refunds, but the net property also subtracts reimbursements. For reimbursed expenses, this changes the reported tag totals and ranking, while daily and category net calculations at L199 and L207 still follow the documented formula.

Content

No source excerpt is available for this finding.