Back to skill

Security audit

Token Connection Guard

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a resource-saving and retry guide, but it also gives under-scoped advice to rotate/revoke API keys and block connections beyond its stated purpose.

Review this skill before installing. It is reasonable for reducing token use, retries, context size, and provider failures, but do not allow it to automatically rotate/revoke keys, regenerate credentials, block connections, or perform audits without explicit user authorization and a clear scope.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Memory PoisoningPersistent Context Injection, Context Window Stuffing, Memory Manipulation
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (10)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The phrase "Use when the conditions in the description are met" does not clearly define concrete activation boundaries or examples, making it unclear when the skill should or should not be invoked. Combined with the broad description in L03, this can cause unintended invocation in many routine situations involving errors or efficiency concerns.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The manifest description at L0006 and the main body describe batching, context compression, retries, backoff, circuit breakers, caching, and provider failover. However, later sections prescribe actions such as rotating leaked tokens, revoking/regenerating API keys, blocking suspicious connections, auditing, and encrypting tokens, which are materially different security operations from the described optimization/stability scope.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The invocation guidance says to use the skill when 'conserving tokens/context/requests' and for handling generic stability issues like timeouts or network errors. These are very common conditions across many tasks, and the file does not provide narrower trigger phrases, scope boundaries, or negative examples in the description field to prevent over-invocation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The operative instructions switch into Indonesian ('Gunakan skill ini ketika', 'Jangan gunakan untuk') and continue that way for most of the file, while the skill does not state that it is Indonesia-specific or provide an opt-in language choice. This creates a natural-language locale policy issue because it effectively constrains operation to a specific language without user consent or justification.

Content

No source excerpt is available for this finding.

Context Window Stuffing

Medium
Category
Memory Poisoning
Confidence
80% confidence
Finding

Skill attempts to fill the context window with filler content, displacing legitimate instructions and safety constraints. This can degrade agent performance or bypass safety boundaries.

Content

Scanner excerpt · SKILL.md (reported line 331)May include surrounding context.

md
AUTH ERROR
SERVER ERROR
MODEL ERROR
CONTEXT OVERFLOW
TOKEN LIMIT
PROVIDER OVERLOAD
ROUTER ERROR

Unbounded Resource Access

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill allows unbounded resource consumption (API calls, storage, compute). Without rate limits or quotas, a compromised or misbehaving agent can cause denial-of-service or cost overruns.

Content

Scanner excerpt · SKILL.md (reported line 1061)May include surrounding context.

md
## 50. NON-NEGOTIABLE RULES

NEVER RETRY FOREVER.

NEVER SEND IDENTICAL FAILED REQUEST REPEATEDLY.

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 · SKILL.md (reported line 1077)May include surrounding context.

md
NEVER IGNORE CONNECTION HEALTH.

NEVER CLAIM SUCCESS WITHOUT VERIFICATION.

NEVER INVENT DATA AFTER NETWORK FAILURE.

Behavior Manipulation

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Subtle instructions detected that may alter agent decision-making or introduce hidden biases.

Content

Scanner excerpt · SKILL.md (reported line 1085)May include surrounding context.

md
ALWAYS USE FAILURE CLASSIFICATION.

ALWAYS PREFER REUSE OVER REGENERATION.

ALWAYS PREFER RELEVANT CONTEXT OVER FULL CONTEXT.

Behavior Manipulation

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Subtle instructions detected that may alter agent decision-making or introduce hidden biases.

Content

Scanner excerpt · SKILL.md (reported line 1087)May include surrounding context.

md
ALWAYS PREFER REUSE OVER REGENERATION.

ALWAYS PREFER RELEVANT CONTEXT OVER FULL CONTEXT.

ALWAYS STOP WHEN SUCCESS IS VERIFIED.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The file overwhelmingly documents a resource-optimization and retry/failover orchestration skill, but the later sections 'Common Mistakes', 'Red Flags', 'Rationalization Prevention', and 'Quick Reference' switch to incident-response guidance about leaked tokens, suspicious connections, revocation, regeneration, and blocking. This is not merely incomplete documentation; it actively changes the apparent intent of the skill from optimization/stability to credential security response.

Content

No source excerpt is available for this finding.