Back to skill

Security audit

password-breach-exposure

Security checks for vulnerabilities and agentic risk

Overview

The assessment can label unchecked passwords as safe and generate misleading rotation priorities, although its network requests follow the advertised privacy approach.

Wait for corrected nickname handling and evidence-based assessment before relying on the rotation queue. Use the local hidden password prompt, keep actual passwords out of chat and inventory files, and protect or delete scripts/data/pwned_cache.json because it retains full password hashes.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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 (7)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

If the implementation behaves as reported, checking password nicknames instead of actual credentials can produce misleading negative results, while inferring breach types from occurrence counts creates unsupported rotation priorities. This is especially dangerous in a credential-triage tool because users may leave compromised passwords unchanged; the implementation claims cannot be independently verified because the script was not supplied.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: file_read, file_write, network.

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.

The email-based assessment command lacks a clear warning about sending the user's email to a breach-lookup service.

Medium
Category
Output Handling
Confidence
86% confidence
Finding

The workflow instructs users to run an assessment with their email address, while lines 67–69 indicate that email/breach checks use separate queries. The privacy section's blanket statement that no email or identifiers are sent obscures this distinction; password-prefix k-anonymity does not protect an email submitted in a separate lookup.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/exposure-playbook.md (reported line 20)May include surrounding context.

md
## K-anonymity flow (exact)

1. `h = SHA1(password.encode('utf-8')).hexdigest().upper()`
2. `GET https://api.pwnedpasswords.com/range/{h[:5]}`
   - No email, no identifiers, no full hash in the URL
   - Optional header `Add-Padding: true` makes even request
     timing/volume less fingerprintable

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · scripts/breach_exposure.py (reported line 30)May include surrounding context.

python
CACHE = DATA / "pwned_cache.json"
CATALOG = DATA / "breach_catalog.json"

API = "https://api.pwnedpasswords.com/range/{}"

# severity weights by leaked credential type
PLAINTEXT, FAST_HASH, STRONG_HASH, CONTACT_ONLY = 100, 80, 65, 10

The rotation queue does not map password hits to actual breach contents; it substitutes corpus-frequency thresholds for leaked credential types.

Medium
Category
Excessive Agency
Confidence
100% confidence
Finding

The assessment assigns plaintext and fast-hash severity weights from password occurrence counts, which do not establish how any breach stored credentials or what information it exposed. Although the catalog is loaded, it does not inform the queue, so the tool presents unsupported breach-severity conclusions that can misdirect credential rotation in a security-sensitive workflow.

Content

No source excerpt is available for this finding.

The documented account inventory accepts password-base nicknames, but assessment checks those nicknames as literal passwords.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The inventory interface describes its second field as a password-base nickname, but assessment submits that nickname to the password checker as though it were the actual password. Consequently, compromised passwords can receive zero-hit, low-priority assessments while unrelated nickname matches can trigger false alarms; this is particularly misleading in a tool intended to prioritize urgent rotations.

Content

No source excerpt is available for this finding.