ClawHub
Back to skill

Security audit

Site Walk Audio Memory

Security checks for vulnerabilities and agentic risk

Overview

This skill is a clear BlueColumn integration for storing and searching construction site-walk notes, with external API use disclosed and aligned with its purpose.

Before installing, confirm BlueColumn is approved for your project data and that its retention, deletion, access-control, and privacy terms fit your organization. Avoid sending regulated, personal, legal-strategy, or highly confidential incident details unless you have approval.

A.I.G

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

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (6)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill instructs users to send site-walk audio, dictated summaries, and query text to BlueColumn's external API, but it does not clearly warn users that potentially sensitive project information will be transmitted to and stored by a third party. In this context, the data can include safety incidents, subcontractor commitments, project status, and possibly other confidential operational details, so omission of an explicit disclosure creates a real privacy and data-governance risk.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
93% confidence
Finding

This example explicitly transmits detailed site-walk content to an external service, including punch items, safety hazards, subcontractor names, and schedule status. The transmission itself is expected for the skill's function, but it is still a genuine security/privacy concern because the content may be confidential and the skill does not provide safeguards, minimization guidance, or user consent language.

Content

Scanner excerpt · SKILL.md (reported line 25)May include surrounding context.

Upload the walk audio, or store a quick dictated summary.

bash
curl -X POST https://api.bluecolumn.ai/remember \
  -H "Authorization: Bearer $BLUECOLUMN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "SITE WALK Oct 11, Tower B level 3: Rojas drywall — SW corner head detail wrong (punch item, due 10/14). Safety: missing guardrail at stair opening (photo taken). Verbal commitment: Rojas will reroute the duct chase before drywall close-in. Elevator core pour not started.", "title": "walk - Tower B L3 Oct 11"}'

External Transmission

Medium
Category
Data Exfiltration
Confidence
93% confidence
Finding

This example explicitly transmits detailed site-walk content to an external service, including punch items, safety hazards, subcontractor names, and schedule status. The transmission itself is expected for the skill's function, but it is still a genuine security/privacy concern because the content may be confidential and the skill does not provide safeguards, minimization guidance, or user consent language.

Content

Scanner excerpt · SKILL.md (reported line 25)May include surrounding context.

Upload the walk audio, or store a quick dictated summary.

bash
curl -X POST https://api.bluecolumn.ai/remember \
  -H "Authorization: Bearer $BLUECOLUMN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "SITE WALK Oct 11, Tower B level 3: Rojas drywall — SW corner head detail wrong (punch item, due 10/14). Safety: missing guardrail at stair opening (photo taken). Verbal commitment: Rojas will reroute the duct chase before drywall close-in. Elevator core pour not started.", "title": "walk - Tower B L3 Oct 11"}'

External Transmission

Medium
Category
Data Exfiltration
Confidence
94% confidence
Finding

The audio upload flow sends a user-provided audio URL for ingestion by an external third party, which can expose sensitive field recordings and associated project context outside the user's environment. In construction workflows, recordings may contain confidential discussions, worker identities, safety issues, or dispute-relevant statements, making the transfer materially sensitive.

Content

Scanner excerpt · SKILL.md (reported line 34)May include surrounding context.

With audio:

bash
curl -X POST https://api.bluecolumn.ai/remember \
  -H "Authorization: Bearer $BLUECOLUMN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"audio_url": "https://your-bucket/walk-towerb-l3-1011.m4a", "title": "walk - Tower B L3 Oct 11 (audio)"}'

External Transmission

Medium
Category
Data Exfiltration
Confidence
91% confidence
Finding

The recall request sends operational queries to an external API, which can reveal project disputes, outstanding subcontractor commitments, and internal status questions even if no raw audio is uploaded. Query content itself may be sensitive business intelligence, and the skill provides no warning or restrictions on what users should ask the service.

Content

Scanner excerpt · SKILL.md (reported line 45)May include surrounding context.

Before a pre-pour meeting or a subcontractor dispute, pull the verbal record.

bash
curl -X POST https://api.bluecolumn.ai/recall \
  -H "Authorization: Bearer $BLUECOLUMN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"q": "What punch items and Rojas commitments are open on Tower B, and when were they made?"}'

External Transmission

Medium
Category
Data Exfiltration
Confidence
92% confidence
Finding

The note endpoint transmits tagged site observations such as safety issues and location details to an external service, creating a record of potentially sensitive incidents and project conditions. Because the skill is designed to centralize these records, the danger is less about malicious behavior and more about undisclosed third-party storage of operationally sensitive data.

Content

Scanner excerpt · SKILL.md (reported line 61)May include surrounding context.

Quick tag for later filtering:

bash
curl -X POST https://api.bluecolumn.ai/note \
  -H "Authorization: Bearer $BLUECOLUMN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"content": "Punch item: stair guardrail missing, Tower B L3, flagged Oct 11 walk", "tags": ["punch-item", "safety", "tower-b"]}'