Back to skill

Security audit

Bytesagain Privacy Policy

Security checks for vulnerabilities and agentic risk

Overview

The skill is advertised as a privacy policy generator, but it also ships an unrelated security CLI that can persist raw user inputs and misrepresents some security functions.

Review this package before installing. The privacy policy template generator is mostly understandable, but the package also includes a separate security-tool script that is not disclosed by the main purpose and can store raw inputs locally. Do not pass secrets, passwords, tokens, or confidential policy details to its commands, and do not rely on its generated policies as legal compliance verification without attorney review.

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 (3)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/script.sh:35
Finding

Plaintext Persistence of Potentially Sensitive Command Inputs

Content
View full analysis
> "$DATA_DIR/history.log"; } cmd_scan() { echo " Scanning for vulnerabilities..." _log "scan" "${1:-}" } cmd_audit() { echo " Running audit checklist..." _log "audit" "${1:-}" } cmd_check() { echo " Checking: $1" _log "check" "${1:-}" } cmd_report() { echo " Security report generated" _log "report" "${1:-}" } cmd_harden() { echo " Step 1: Update | Step 2: Firewall | Step 3: Auth" _log "harden" "${1:-}" } cmd_encrypt() { echo " Encrypting: $1" _log "encrypt" "${1:-}" } cmd_hash() { echo "$1" | sha256sum | cut -d" " -f1 _log "hash" "${1:-}" } ``` ### Technical Analysis The shared `_log` function writes command arguments directly into the persistent `$DATA_DIR/history.log` file. In particular, the `encrypt` and `hash` commands accept values that users may reasonably expect to be secrets, credentials, tokens, personal information, or cryptographic material. Although `hash` produces a SHA-256 digest on standard output, it subsequently records the original unhashed input. The `encrypt` command similarly records its original input without protecting it. The script does not explicitly create the data directory and history file with restrictive permissions, does not redact sensitive values, and does not implement retention or cleanup. This creates a confidentiality flaw rather than a privilege-escalation vulnerability. Exploitation requires filesystem access to the resulting log, such as access through another local account, a compromised process, an insecure backup, or accidentally broad permissions. ### Attack Path 1. A user invokes `scripts/script.sh hash ` or `scripts/script.sh encrypt `. 2. The command passes the orig ...[truncated 856 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/script.sh:71
Finding

Password Generator Uses a Non-Cryptographic Pseudorandom Number Generator

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/script.sh:58
Finding

Encryption Command Misrepresents Its Behavior and Logs Plaintext Instead of Encrypting It

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (14)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The skill description claims it can scan data practices, audit consent flows, and check compliance, but the analyzed behavior appears limited to static policy generation rather than real assessment or validation. Users could rely on the skill for legal or compliance decisions it is not capable of making, producing false assurance and potentially causing regulatory, contractual, or privacy failures.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill description claims it can scan data practices, audit consent flows, and check compliance, but the analyzed behavior appears limited to static policy generation rather than real assessment or validation. Users could rely on the skill for legal or compliance decisions it is not capable of making, producing false assurance and potentially causing regulatory, contractual, or privacy failures.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The script's implemented behavior materially differs from the declared skill purpose: instead of generating GDPR/CCPA privacy policies, it exposes a generic security scanning and hardening CLI. This kind of capability mismatch is dangerous because it breaks user trust, can enable unexpected data handling or operational behavior, and may conceal functionality the user did not intend to invoke in a privacy-policy skill context.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The inline documentation explicitly labels the tool as a 'Security scanning and hardening tool,' directly contradicting the published privacy-policy metadata. Such contradictions are a strong indicator of deceptive or poorly governed packaging, and in an agent ecosystem they increase the risk that operators will grant trust or permissions based on false expectations.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
85% confidence
Finding

The skill advertises executable capabilities and the analyzer detected network-related behavior, but the manifest declares no explicit tool scope such as permissions or allowed-tools. This creates an authorization ambiguity where a host agent may grant broader access than users expect, reducing transparency and making misuse or data egress harder to govern.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The broad 'Use when...' activation text covers scanning, auditing, compliance checking, and drafting without clear boundaries or exclusions. In agent environments, vague triggers increase the chance that the skill is auto-selected in contexts beyond its real competence, which can lead to overcollection of information, misleading outputs, or inappropriate workflow decisions.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Generic single-word commands like 'generate', 'gdpr', 'ccpa', 'app', 'website', and 'audit' are likely to collide with ordinary conversation or other skills. In an agentic system this raises the risk of unintended invocation, causing the skill to activate on ambiguous user text and process data or produce outputs the user did not explicitly request.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The script's user-facing help and generated documents include mandatory Chinese alongside English, such as '隐私政策生成器' and bilingual section headings, with no option to select language or opt out. This creates a language/locale policy concern because the skill forces a specific locale presentation on all users rather than offering a choice or clearly justifying the requirement.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% 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 · scripts/privacy.sh (reported line 472)May include surrounding context.

sh
We use cookies in compliance with the ePrivacy Directive. Users are
presented with a cookie consent banner upon first visit. Categories:

  • Strictly Necessary: Always active (no consent required)
  • Analytics: Require consent
  • Marketing: Require consent
  • Preference: Require consent

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

Hashing and password-generation utilities are unrelated to drafting or auditing privacy policies and represent unnecessary capability expansion for this skill. In context, this is dangerous because extra features increase attack surface, can be abused by downstream workflows, and reinforce that the package is not operating within its declared scope.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The script logs raw command arguments to a history file without notice, which can capture sensitive user-supplied content such as compliance details, file paths, secrets, or policy text. In a privacy-focused skill context, silent retention of user inputs is especially problematic because it undermines confidentiality expectations and may create unauthorized local data storage.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The file presents key usage text and tips in both Chinese and English, but does not state whether the skill will respond in one language, both, or based on user preference. This can violate language-choice expectations if the skill defaults to a language or locale without explicit opt-in.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

The documented 'Analytics' command is unrelated to the stated privacy-policy generation purpose, which can confuse invocation boundaries and create an unexpected feature surface. Even if benign, undocumented or mismatched capabilities increase the chance of inappropriate activation, data handling surprises, or hidden functionality concerns.

Content

No source excerpt is available for this finding.

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

Scanner excerpt · scripts/privacy.sh (reported line 94)May include surrounding context.

sh
1.1 Personal Information You Provide

  When you use our services, we may ask you to provide certain personally
  identifiable information, including but not limited to:

  • Full name
  • Email address (${EMAIL} for privacy inquiries)

Static analysis

No suspicious patterns detected.