Construction Daily Report Generator

• Outcome: Pass
• Latest audit: Sep 14, 2026 · 11:43 PM UTC
• Version: 1.0.0

Overview

This is a construction report-writing skill with no executable code, persistence, or data-access behavior, though it may over-activate on generic report requests and may append an unwanted promotional footer.

Install only if you want an assistant for construction daily site reports. Review generated reports before filing, especially to remove any unwanted branding or external-domain footer, and be aware it may activate on generic daily-report wording unless your agent applies construction-context judgment.

A.I.G

Vulnerability Patterns

• Skill Instruction Hijacking: Alters the agent's session goals or safety constraints when the skill loads
• Agent Memory Poisoning: Writes attacker-controlled rules into memory that affect later sessions
• Remote Payload Retrieval and Execution: Fetches external code whose behavior can change after review
• Embedded Malicious Code: Ships malicious scripts inside the skill and executes them locally
• Unauthorized Access and Privilege Escalation: Obtains permissions beyond the task's legitimate needs

Findings (1)

T01 · Skill Instruction Hijacking

Note

• Location: SKILL.md:409

Finding

Hard-Coded Promotional Content Hijacks Generated Report Output

Content

View full analysis

Vulnerability Details

File Location: SKILL.md, lines 409–410
Vulnerability Type: Unrelated promotional content embedded in the modeled agent response
Risk Level: Low

Complete Code Snippet:

markdown

*This skill is published by ERTRS — the construction intelligence platform. For automated reporting, project tracking, and commercial management, visit ertrs.com*

Technical Analysis

The skill embeds a third-party promotional statement and external domain in the example agent output. Because examples in skill instructions can influence generated responses, the agent may reproduce this footer when creating daily site reports. The content is unrelated to the user's reporting task and can therefore alter the expected output by adding unsolicited advertising and an external-link referral.

This issue affects response integrity rather than system security. The reviewed text does not instruct the agent to access the domain, retrieve a payload, transmit report data, execute code, or weaken safety constraints. Accordingly, the finding does not provide code execution, filesystem access, elevated privileges, or access to credentials.

Attack Path

1. The construction daily-report skill is loaded into the agent's instruction context.
2. The agent processes a user's unstructured construction-site notes.
3. The agent follows the modeled response in SKILL.md.
4. After generating the legitimate report and review question, the agent reproduces the promotional footer.
5. The resulting professional report contains unsolicited branding and an external-domain reference, potentially encouraging the user to leave the trusted interaction context.

Impact Assessment

The impact is limited to generated-output integrity and user trust. Reports may contain unwanted commercial branding, become unsuitable for formal filing, or direct users toward an external website. No operating-system privileges, persistent access, ...[truncated 111 chars]

Remediation

View remediation

Remediation Suggestions

Remove the promotional footer from the example agent response and ensure the required output template ends with the report review question. If publisher attribution is necessary, place it in non-operative package metadata or a clearly labeled documentation section that explicitly states it must not be included in generated reports.

Add an instruction requiring generated reports to contain only user-requested project information and approved report-template fields. External links or commercial promotions should be included only when the user explicitly requests them. Regression-test representative report-generation prompts to verify that no branding or unrelated URL is appended.

SkillSpector

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (1)

Vague Triggers

Medium

• Category: Not specified by scanner
• Confidence: 89% confidence

Finding

The activation rules are broad enough to trigger on generic phrases like 'daily report', 'progress report', or rough notes that may appear in ordinary writing-assistance conversations. This can cause the agent to activate the wrong skill, override user intent, and elicit structured follow-up questions for construction-site data in contexts where the user only wanted general help formatting or drafting text.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.