Back to skill

Security audit

Finance Accrual Rollforward Coordinator

Security checks for vulnerabilities and agentic risk

Overview

This skill is a small finance workflow helper whose memory and file-use instructions are disclosed and aligned with accrual rollforward coordination.

Before installing, be comfortable with the skill using saved agent memory to reuse finance workflow locations and conventions, and with it editing local reconciliation or issue-tracking files when the active task requires that work. Review saved memory contents if those paths or ownership conventions may be sensitive across teams or workspaces.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • 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 (1)

T05 · Unauthorized Access and Privilege Escalation

Warning
Location
SKILL.md:11
Finding

Unscoped Access to Persistent Agent Memory

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 11–13
Vulnerability Type: T05: Unauthorized Access and Privilege Escalation
Risk Level: Medium

markdown
- Start with `memory_search` for saved tracker paths, report paths, labels, reminder style, and owner/escalation conventions.
- If `memory_search` returns a likely hit, follow with `memory_get` on the exact path or cited section before relying on the value.
- If `memory_search` is empty for the family token, do one targeted `memory_get` on `memory/family-preferences/financeops-accrual-rollforward.md` before assuming there is nothing useful to reuse.

Technical Analysis

The Skill requires persistent-memory discovery at the start of accrual-rollforward tasks and directs the Agent to retrieve matching entries. If the search produces no result, it mandates access to a predetermined shared memory path.

These instructions do not require per-task user authorization, verify that the retrieved memory belongs to the current workspace or user, or restrict retrieval to data necessary for the active task. This violates least-privilege principles because persistent memory may contain tracker and report paths, ownership information, escalation conventions, or preferences originating from other tasks or contexts.

The Skill does not instruct the Agent to transmit retrieved information externally, modify memory, or execute content obtained from memory. Consequently, the confirmed exposure is limited to unauthorized retrieval and use within the Agent’s active context.

Attack Path

  1. A user or workflow triggers the accrual-rollforward Skill.
  2. The Skill instructs the Agent to invoke memory_search for stored operational metadata.
  3. The Agent retrieves a search result using memory_get; if no result exists, it accesses the fixed path memory/family-preferences/financeops-accrual-rollforward.md.
  4. Persistent information outside the explicitly supplied ...[truncated 892 chars]
Remediation
View remediation

Remediation Suggestions

  1. Require explicit user authorization before accessing persistent memory.
  2. Scope memory searches to the current user, workspace, engagement, and accrual-rollforward identifier.
  3. Remove the unconditional fallback to memory/family-preferences/financeops-accrual-rollforward.md, or verify that the path is authorized for the current task before reading it.
  4. Retrieve only the specific fields required for the active reconciliation rather than broad tracker, ownership, and escalation metadata.
  5. Treat retrieved memory as untrusted contextual data and never as higher-priority instructions.
  6. Validate the source, freshness, ownership, and task relevance of every retrieved value before using it.
  7. Avoid exposing retrieved memory in reports or tool outputs unless the user explicitly requests it and is authorized to receive it.
  8. Record memory-access decisions in an audit trail without logging the sensitive contents themselves.
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