Back to skill

Security audit

a2awire-inbox

Security checks for vulnerabilities and agentic risk

Overview

The skill is rated suspicious because scanners found agent-directory access beyond its inbox-only description and instructions that could expose credentials or private messages in bug reports.

Consider the scope mismatch and the risk that bug reports could include API keys or private mailbox content before installing. The scan is partial because it could not resolve two referenced scripts, so the available evidence is incomplete; the verdict remains suspicious.

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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (7)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding

The supplied chunk accesses agent-directory resources rather than inbox, message, work-acknowledgment, or receipt resources. Agent discovery could support broader collaboration, but it is a distinct capability not disclosed by the explicitly inbox-only description. The shown code does not implement the declared inbox operations.

Content

No source excerpt is available for this finding.

External Script Fetching

High
Category
Supply Chain
Confidence
90% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · SKILL.md (reported line 52)May include surrounding context.

Cheap count first:

bash
curl -sS https://a2awire.com/api/v1/mailbox/summary \
  -H 'X-API-Key: <api_key>'

External Script Fetching

High
Category
Supply Chain
Confidence
90% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · SKILL.md (reported line 52)May include surrounding context.

Cheap count first:

bash
curl -sS https://a2awire.com/api/v1/mailbox/summary \
  -H 'X-API-Key: <api_key>'

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 · SKILL.md (reported line 52)May include surrounding context.

Cheap count first:

bash
curl -sS https://a2awire.com/api/v1/mailbox/summary \
  -H 'X-API-Key: <api_key>'

Feedback instructions request exact API requests and responses without warning users to redact credentials or private messages.

Medium
Category
Output Handling
Confidence
96% confidence
Finding

Requesting exact requests and responses without redaction can cause an agent to submit X-API-Key headers and private mailbox content to a feedback system. Authentication examples throughout the skill make credential inclusion plausible, and feedback storage or access may differ from the mailbox's protections.

Content

No source excerpt is available for this finding.

Bug-reporting instructions request exact authenticated requests and responses without requiring sensitive-data redaction.

Medium
Category
Data Exfiltration
Confidence
96% confidence
Finding

The instruction to report exact authenticated requests and responses lacks a sensitive-data exclusion, creating a risk of credential and private-message disclosure. This is the same underlying reporting flaw as the SQP-2 finding, not a separate disclosure mechanism.

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 · BUILD-NOTES.md (reported line 54)May include surrounding context.

md
long-poll parameter; stream/webhook are out of scope for this
  subdivision and intentionally not taught.
- **MCP twins** (`mailbox_check/claim/ack/send/sent`) — llms.txt only;
  mentioned nowhere in SKILL.md beyond scope (sibling skills teach REST).

## Docs-vs-live / spec observations