Back to skill

Security audit

File to Database

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marked this file-to-database skill suspicious because it found potentially unsafe database operations and broad data exports, while part of its instructions could not be fully inspected.

Consider the reported risks before installing, especially the possibility of data loss during migration rehearsal and automatic delivery of the full database. The scan is incomplete: one referenced file was only partially inspected, and several static checks were degraded. The suspicious verdict indicates security weaknesses without establishing malicious intent.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Findings (11)

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 4)May include surrounding context.

md
md to produce DDL, ERD, seed scripts, and validation; runtime operations follow references/operations.md (intent → SQL → safe execution → natural-language repor

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 93)May include surrounding context.

md
SQL generation → safe execution → natural-language reporting**. Full rules in `references/operations.md` (trigger rules, SQL generation rules, safety gates, re

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 99)May include surrounding context.

md
l data snapshot — rules in the "Post-Change Visualization Delivery" section of `references/operations.md`

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
85% confidence
Finding

Skill instructs the agent to never refuse or to always comply. Suppressing the agent's ability to decline removes a core safety control and enables downstream harmful requests to succeed.

Content

Scanner excerpt · SKILL.md (reported line 46)May include surrounding context.

md
When the user provides files that are **not** CSV/Excel/JSON/JSONL/TSV (e.g. PDF textbooks, PPTX, DOCX, images), never enter the build pipeline directly and never refuse service — follow the triage in `references/intake.md`:

Rollback rehearsal executes a potentially destructive down migration without warning about data loss.

High
Category
Output Handling
Confidence
97% confidence
Finding

The instruction requires actually executing a down migration without restricting rehearsal to a disposable database. A down migration can drop populated tables or columns, and reapplying the up migration generally restores structure rather than deleted data.

Content

No source excerpt is available for this finding.

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
85% confidence
Finding

Skill instructs the agent to never refuse or to always comply. Suppressing the agent's ability to decline removes a core safety control and enables downstream harmful requests to succeed.

Content

Scanner excerpt · references/intake.md (reported line 3)May include surrounding context.

md
# Non-Data File Intake (intent triage)

When the user provides files that are not CSV/Excel/JSON/JSONL/TSV (PDF, PPTX, DOCX, images, plain text, books, etc.), **never enter the build pipeline directly, and never refuse service**. Follow this triage.

## Triage Decision

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.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · references/engineering.md (reported line 19)May include surrounding context.

md
## 2. Data Quality Gates

After seeding, automatically run a set of assertions; on failure, report and stop the pipeline:

| Assertion | Default threshold |
|---|---|

The migration workflow directs dropping old columns without warning about permanent data loss.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

The expand-and-contract workflow explicitly includes dropping existing columns, but does not warn that this can permanently remove their stored data. Requiring a rollback script does not ensure that deleted values can be recovered.

Content

No source excerpt is available for this finding.

Automatic delivery exports database contents without a privacy warning.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

Every modification automatically triggers export and delivery of the entire database, even when the user requested only a narrow record change. This can expose unrelated confidential records through chat attachments or other delivery channels; snapshot masking and PII reminders elsewhere do not establish authorization to transmit the full database.

Content

No source excerpt is available for this finding.

Foreign keys can receive an automatic-adoption verdict despite failing the documented 100% containment requirement.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

For an exact name match, containment of 0.99 receives confidence 0.8 and an automatic-adoption verdict despite the documented requirement for complete containment. The comparison also uses samples rather than complete value sets, allowing invalid relationships to be adopted and causing import failures or incorrect schema constraints.

Content

No source excerpt is available for this finding.