Back to skill

Security audit

Geo Push Policy

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent notification-policy helper with local state storage and retry logic, with no evidence of hidden control, exfiltration, or destructive behavior.

Before installing, make sure a Chinese-language notification policy helper fits your workflow. Use a limited local state-file path, and review the cooldown and push-count logic before relying on this for important alerts because it may duplicate or suppress notifications. Failed messages can be retained in the local dead-letter queue and resent through the caller-provided send function.

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 (2)

T09 · Insecure Skill Coding Practices

Warning
Location
geo_push_policy.py:101
Finding

Maximum Push Limit and Cooldown Can Be Bypassed by Reusing a High-Scoring Event

Content
View full analysis
= policy.threshold_high and push_count > 0: return True, f"Event upgrade during cooldown, score {total_score}" return False, f"Within cooldown ({elapsed_minutes:.0f}/{cooldown} minutes)" except Exception as e: print(f"Failed to parse time: {e}") # Check the number of pushes if push_count >= policy.max_push_count: return False, "Maximum push count reached" ``` ### Technical Analysis The function returns a successful push decision during the cooldown whenever both of the following conditions are true: 1. The current score is at least the high-priority threshold. 2. The event has previously been pushed at least once. It does not verify that the event actually transitioned from a lower severity to high severity. Consequently, an event that was already high severity is repeatedly treated as an upgrade. In addition, the maximum-push-count check occurs after this successful return. An event whose `push_count` is already equal to or greater than `policy.max_push_count` can therefore still be ap ...[truncated 1508 chars]
Remediation
View remediation
= policy.max_push_count: return False, "Maximum push count reached" ``` 2. Replace the score-only cooldown exception with a genuine state transition check: ```python if elapsed_minutes < cooldown: if check_event_upgrade(event_dict, state): return True, f"Event upgraded during cooldown, score {total_score}" return False, f"Within cooldown ({elapsed_minutes:.0f}/{cooldown} minutes)" ``` 3. Record the previous score and severity explicitly, and define upgrade conditions as transitions rather than static threshold checks. 4. Apply the push-count check again in the delivery/update layer so a caller cannot bypass the policy by submitting an event directly to the push list. 5. Add regression tests covering: - An already-high event during cooldown. - An event with exactly three prior pushes. - An event with more than three prior pushes. - A genuine transition from medium to high priority. - Repeated calls with an unchanged high score. ]]>

T09 · Insecure Skill Coding Practices

Note
Location
geo_push_policy.py:101
Finding

Cooldown Is Reset by Event Observation Instead of Successful Push Delivery

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (2)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The natural-language content of the skill is presented exclusively in Chinese, including the title, feature descriptions, and usage guidance. Under the stated policy, forcing a specific language without user opt-in is a language/locale policy violation unless the restriction is explicitly justified, which is not present here.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The module name and descriptive docstring are written entirely in Chinese, which imposes a specific language/locale in the skill's natural-language interface. There is no indication that the skill is region-specific or that users can opt into another language, which matches the locale-policy violation criteria.

Content

No source excerpt is available for this finding.