Back to skill

Security audit

huawei-cloud-doris-sql-check

Security checks for vulnerabilities and agentic risk

Overview

This skill checks SQL locally without database access, but its validation can incorrectly report incomplete statements as passing.

Use this as an offline advisory linter, not as authorization to execute SQL: a passing report does not establish valid or safe SQL. Reports reproduce the submitted SQL. Cluster credentials are unnecessary; any separately requested live profiling should have explicit authorization and database-side limits. Independent inspection addressed the scanner's material coverage gaps, but comprehensive grammar correctness was not established.

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
  • Memory PoisoningPersistent Context Injection, Context Window Stuffing, Memory Manipulation
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 336)May include surrounding context.

md
| [Syntax Rules](rules/syntax_rules.yaml) | 34 syntax check rule definitions |

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 337)May include surrounding context.

md
| [Specification Rules](rules/spec_rules.yaml) | 40 specification check rule definitions |

Memory Manipulation

High
Category
Memory Poisoning
Confidence
80% confidence
Finding

Skill manipulates agent memory, state, or stored context. Memory corruption can alter personality, override safety rules, or cause unpredictable behavior.

Content

Scanner excerpt · scripts/doris_sql_parser.py (reported line 487)May include surrounding context.

python
)

    def _parse_delete(self):
        """Parse DELETE statement"""
        self._expect_keyword("DELETE")
        self._expect_keyword("FROM")
        table = self._consume_qualified_name()

PERF005 configures execution-based database profiling beyond the advertised syntax and specification checks.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

PERF005 requests execution-based profiling through MCP and applies it to SELECT, UPDATE, and DELETE without specifying authorization, isolation, or resource limits. Unlike offline syntax checking, execution-based analysis can consume production resources and, depending on database support and execution semantics, risk modifying data; the supplied configuration does not establish that such execution actually occurs.

Content

No source excerpt is available for this finding.

Claimed comprehensive syntax validation falls back to accepting tokens without checking statement grammar.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The generic fallback consumes tokens without validating statement grammar or recording an unsupported-validation error, while is_valid() treats the absence of errors as success. Consequently, recognized statement types without dedicated parsers can receive a misleading valid result, undermining the advertised SQL review and potentially bypassing downstream checks that trust this result; this code does not itself execute SQL.

Content

No source excerpt is available for this finding.