Back to skill

Security audit

Nuclear Accident Human Lifeline

Security checks for vulnerabilities and agentic risk

Overview

This skill has no executable payload, but it asks an agent to keep persistent files with sensitive details about nearby people and emergency resources without clear privacy safeguards.

Review before installing. Use it only for explicit preparedness planning, get permission before recording anyone's contact details or resources, minimize or redact sensitive fields, store any records in a restricted or encrypted local location, avoid cloud sync or source control, and follow official emergency instructions first during a real incident.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (2)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:29
Finding

Plaintext Storage of Sensitive Third-Party and Preparedness Data

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 29-49
Vulnerability Type: Plaintext sensitive-data storage and unsafe filename construction
Risk Level: Medium

Vulnerable Code Snippet:

markdown
**Agent action**: Create a file `nuclear-map.md` with columns: Name/Contact, Asset/Skill, Distance, Trust Level (low/medium/high), Last Contact Date. Fill it only with people you can reach organically — no online strangers.

**Agent action**: Use this exact opener script: “With everything going on in the world lately, I’ve been thinking about how handy it would be to know a few neighbors who could check on each other if something big ever happened. You seem like someone who’s prepared — ever thought about that?” Listen more than you speak. Log every interaction in `contact-[name].md`.

**Agent action**: Use this script: “If anything serious ever hits — power outage, storm, or worse — would you be open to a quick group text or knock to share what we each know and maybe use each other’s basements or supplies? I’d do the same for you.” Get verbal agreement and exchange emergency phone numbers. Note the agreement date in their file.

**Agent action**: Open `nuclear-map.md`, update trust levels and assets, and schedule one low-key check-in per month with at least two people.

Technical Analysis

The Skill requires an agent to persist names, contact details, approximate locations, trust assessments, emergency capabilities, preparedness assets, interaction histories, and agreement dates in predictable Markdown files. These instructions do not specify encryption, restrictive file permissions, data minimization, consent documentation, retention limits, secure deletion, or exclusion from backups and source control.

The dynamic filename pattern contact-[name].md also lacks filename sanitization requirements. If a name or identifier contains path separators or traversal sequences, an implementation that performs direc ...[truncated 1855 chars]

Remediation
View remediation

Remediation Suggestions

  • Obtain explicit, informed consent before recording any third party's personal or preparedness information.
  • Minimize stored data. Avoid detailed supply inventories, subjective trust ratings, precise locations, and complete phone numbers unless strictly necessary.
  • Store sensitive records in an encrypted location with permissions restricted to the current user.
  • Prevent automatic inclusion in source control, cloud synchronization, telemetry, logs, and unencrypted backups.
  • Define retention periods and provide procedures for reviewing, correcting, exporting, and securely deleting records.
  • Replace names in filenames with random identifiers or sanitized, application-generated identifiers.
  • Reject path separators, traversal sequences, control characters, reserved names, and unexpected filename characters.
  • Resolve the final path and verify that it remains inside a dedicated storage directory before writing.
  • Document that verbal participation in an emergency pact does not automatically constitute consent to maintain a detailed digital profile.

T08 · Insecure Dependencies

Warning
Location
SKILL.md:11
Finding

Unpinned Package Execution Through npx

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, line 11
Vulnerability Type: Unpinned executable dependency and mutable supply-chain source
Risk Level: Medium

Vulnerable Code Snippet:

yaml
install: "npx clawhub install nuclear-accident-human-lifeline"

Technical Analysis

The installation command invokes clawhub through npx without pinning the CLI package to an audited version or integrity digest. If an appropriate trusted local copy is unavailable, npx may resolve and execute a package from the configured package registry. The effective executable can therefore differ from the version originally reviewed.

The project contains no lockfile, vendored installer, checksum, signature requirement, or registry restriction that establishes which CLI implementation must execute. This creates a supply-chain trust boundary at installation time. A compromised package release, registry account, registry configuration, or dependency of the resolved package could introduce code not represented in the audited project.

Attack Path

  1. A user follows the metadata installation instruction.
  2. npx resolves the unpinned clawhub package using the user's configured package source or an available local cache.
  3. An attacker compromises the package, a transitive dependency, its publisher account, or the configured registry and supplies a modified release.
  4. npx downloads and executes the modified package.
  5. The malicious installer runs with the same filesystem, network, environment-variable, and process privileges as the user who invoked the command.
  6. It can perform actions within that user's permission scope before or while presenting expected installation behavior.

Impact Assessment

A compromised package could execute arbitrary code with the invoking user's privileges. Depending on that account's permissions and environment, the accessible scope could include project files, user-owned fil ...[truncated 478 chars]

Remediation
View remediation

Remediation Suggestions

  • Pin the CLI to a specific audited version, for example by using an explicitly versioned package reference.
  • Use a lockfile and verify package integrity metadata before execution.
  • Restrict installation to an approved registry and document the expected package publisher and source.
  • Prefer a previously installed, organization-managed CLI over implicit download and execution through npx.
  • Verify package signatures or published checksums where supported.
  • Audit the pinned package and its transitive dependencies before approving upgrades.
  • Execute installation with a non-privileged account in a constrained environment with minimal filesystem, credential, and network access.
  • Avoid exposing unrelated secrets or authentication tokens to the installer process.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (4)

Rp1

Medium
Category
MCP Rug Pull
Confidence
96% confidence
Finding

The install command invokes npx clawhub without pinning a specific version, which creates a supply-chain risk because the resolved package version can change over time or be compromised upstream. In a skill ecosystem, this is especially risky because users may run the install command verbatim, trusting the skill metadata to fetch and execute code.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The 'When to Use' section is broad enough to trigger the skill in loosely related or high-stress situations, including active emergencies, which increases the chance an agent will apply the workflow without clear user consent or appropriateness checks. Because the skill directs relationship mapping, contact logging, and activation behavior, ambiguous invocation can lead to unnecessary collection and use of sensitive personal information.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This instruction tells the agent to create nuclear-map.md containing names, contact details, trust levels, and proximity data, which is sensitive personal information and potentially a targeting or safety risk if exposed. The context makes it more dangerous because the file centralizes a local resilience network and labels individuals by assets, making the information valuable for misuse even if the skill's overall purpose is benign.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Logging interpersonal interactions, verbal agreements, and emergency phone numbers creates a structured dossier of private relationship data without any stated safeguards, consent requirements, or handling limits. In this skill's context, the data is particularly sensitive because it links identities, preparedness status, shelter options, and emergency coordination details that could be abused if accessed by others.

Content

No source excerpt is available for this finding.