Back to skill

Security audit

CFO / Chief Financial Officer

Security checks for vulnerabilities and agentic risk

Overview

This CFO skill is a disclosed advisory finance guide with limited local preference storage and no executable code, network access, bank connections, or money movement.

Before installing, be comfortable with the skill reading and maintaining local CFO preference/context files under ~/Clawic/data/cfo/ and reading shared profile values when needed. Do not store credentials, account numbers, tax IDs, raw bank/payroll exports, or sensitive employee/customer detail in its memory files.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Hidden Instructions

High
Category
Prompt Injection
Confidence
70% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · memory-template.md (reported line 13)May include surrounding context.

md
last: YYYY-MM-DD

## Company Shape
<!-- Stage, headcount range, business model, last round if mentioned -->

## Finance Stack
<!-- Accounting, billing, payroll, expense, cap table systems in use -->

Hidden Instructions

High
Category
Prompt Injection
Confidence
70% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · memory-template.md (reported line 17)May include surrounding context.

md
## Finance Stack
<!-- Accounting, billing, payroll, expense, cap table systems in use -->
<!-- Auditor, tax advisor, banks — names only, never credentials or account numbers -->

## Calendar
<!-- Fiscal year end, close timing, board meeting months, investor update date, budget kickoff -->

Unbounded Output

Medium
Category
Output Handling
Confidence
60% confidence
Finding

Output size or generation rate is not bounded. Unbounded output enables denial-of-service through resource exhaustion, log flooding, or context-window stuffing.

Content

Scanner excerpt · cost-cuts.md (reported line 61)May include surrounding context.

md
- Say the real thing: "We are extending runway and reviewing every contract. I would rather keep you at a lower number than replace you." Renewal dates are leverage; ask 60 days out, not on the renewal date.
- Cloud and infrastructure: right-size before renegotiating a commitment, or you lock in the waste for another year.

## Buying Time Without Cutting

Ranked by cost, cheapest first: collections push and early-pay incentives on the largest overdue balances → vendor term extension → an undrawn facility already in place → annual-prepay offers to existing customers → an insider bridge note → new debt → a structured down round. Structure taken under duress is deferred dilution with interest; price it as such.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill explicitly instructs the agent to read from and write to user-local files under ~/Clawic/data and ~/Clawic/profile.yaml without any explicit user notice or consent step in this setup file. Even though the stated purpose is preference loading and limited persistence, this creates a privacy and integrity risk because sensitive local context may be accessed or modified automatically, and the skill also persists conversation-derived data across sessions.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

This markdown file includes a natural-language trigger phrase, 'Trigger to hedge at all,' followed by a heuristic threshold based on 'one month of net burn.' The condition is somewhat underspecified because it does not define how net burn should be measured or note exceptions, which could lead to inconsistent or overly broad application of the guidance.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.