Back to skill

Security audit

lgpd-audit-logging

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious because scanners report weaknesses that could make its audit trail unreliable or allow records to be altered without detection.

Consider postponing installation until the integrity protections and concurrent-write behavior have been reviewed and corrected. Skillspector's analysis was partial: it reports only 33.3% coverage, with two files partly inspected, so the scan cannot establish that unreported behavior is safe.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
assets/audit-log.ts
Finding

Unhashed audit fields can be altered without invalidating the chain or external anchor

Content
View full analysis

Impact

An attacker can falsify recorded legal justification, purpose, client IP, or user agent without changing any chain hashes. Verification still returns valid, even when an external anchor exists. This undermines the documented claim that edits to any column are detected.

Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

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 29)May include surrounding context.

md
- **Implementação TypeScript serializada + verificador**: [`assets/audit-log.ts`](assets/audit-log.ts)

The writer promises transactionally serialized appends but silently proceeds when its locking capability is absent.

High
Category
Excessive Agency
Confidence
99% confidence
Finding

The transaction interface makes the advisory-lock capability optional, and the writer silently continues when it is unavailable. Without equivalent database serialization, concurrent transactions can read the same chain head and append entries with identical prevHash values, producing a fork that verification subsequently rejects. This undermines the skill's explicit concurrency-safety guarantee and can make the audit trail unreliable under ordinary concurrent traffic or deliberately induced load; adapters that provide the lock are not exposed to this particular fallback.

Content

No source excerpt is available for this finding.

The verifier claims to detect edits to any column, but several stored fields are not integrity-protected.

High
Category
Excessive Agency
Confidence
100% confidence
Finding

The self-hash excludes stored fields including legalBasis, purpose, ipAddress, userAgent, and id, so changes to these fields pass verification without changing any chain hashes or triggering an external anchor mismatch. An attacker able to modify stored records could therefore falsify compliance context or attribution while the verifier reports a valid chain. This is especially dangerous in a skill specifically promising tamper-evident LGPD audit evidence, although effective database write protections could limit exploitation.

Content

No source excerpt is available for this finding.