Back to skill

Security audit

授信报告全流程

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly matches a fixed-template credit-report workflow, but it tells the agent to keep updating the installed skill based on project experience, which is risky for sensitive financial work.

Install only if you are comfortable with a workflow that handles sensitive borrower financial materials. Before use, remove or gate the continuous-update instruction so changes to the installed skill require explicit human review and do not persist customer-derived details.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Rogue AgentSelf-Modification, Session Persistence
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (6)

Self-Modification

High
Category
Rogue Agent
Confidence
92% confidence
Finding

The skill instructs the agent to 'Update this skill whenever a project reveals' new patterns, which is a self-modification directive. In an adversarial or mixed-trust environment, that can let untrusted project content influence persistent skill behavior, causing prompt injection persistence, policy drift, or incorporation of malicious instructions into future runs. The credit-report context lowers direct system impact compared with code-execution skills, but the persistence risk is still real because future analyses or document workflows could inherit tainted guidance.

Content

Scanner excerpt · SKILL.md (reported line 395)May include surrounding context.

md
## Continuous Update Guidance

Update this skill whenever a project reveals:
- a new template pitfall
- a new stale-value pattern
- a stronger examiner objection pattern

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 · SKILL.md (reported line 23)May include surrounding context.

md
- Free-form narrative reports without a fixed template
- Pure legal opinions, audit reports, or valuation reports
- Tasks limited to a single isolated artifact when a narrower skill is enough
  - Use [$spreadsheet](/Users/daisy/.codex/skills/spreadsheet/SKILL.md) for spreadsheet-only work
  - Use [$credit-report-writing](/Users/daisy/.codex/skills/credit-report-writing/SKILL.md) for text-only credit writing

## Required Inputs

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 · SKILL.md (reported line 24)May include surrounding context.

md
- Pure legal opinions, audit reports, or valuation reports
- Tasks limited to a single isolated artifact when a narrower skill is enough
  - Use [$spreadsheet](/Users/daisy/.codex/skills/spreadsheet/SKILL.md) for spreadsheet-only work
  - Use [$credit-report-writing](/Users/daisy/.codex/skills/credit-report-writing/SKILL.md) for text-only credit writing

## Required Inputs

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

The instruction to write 'approval-oriented paragraphs' encodes a specific organizational/locale communication style as mandatory output behavior. Because the skill does not offer a user opt-in or explain that this is a justified region-specific compliance requirement, it can be read as forcing a particular language/register policy.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill enables implicit invocation but does not define clear trigger constraints or narrowing conditions. That can cause the agent to auto-select this skill in ambiguous contexts, potentially exposing sensitive financial documents or causing unauthorized drafting/review actions in a bank-credit workflow where data is highly sensitive and consequential.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

The metadata uses Chinese-only natural-language descriptions and keywords, which suggests the skill is intended to operate in a specific language without stating that users can opt in or choose another locale. Under the policy, forcing a specific language without user choice or documented justification is a natural-language policy concern.

Content

No source excerpt is available for this finding.