Security audit

Everwood AIDLC (OpenClaw)

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent planning workflow, but one helper can copy any readable local text file into persistent session artifacts if invoked with an unsafe path.

Review this before installing if you work in repositories with secrets or sensitive customer data. Only invoke gate-lock with gate files inside the active aidlc-sessions/<uuid>/gates directory, avoid putting secrets in gate artifacts, and consider patching the helper to reject paths outside the current session before broad use.

A.I.G

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 (1)

T09 · Insecure Skill Coding Practices

Warning

Location: scripts/gate-lock.py:63
Finding: Unrestricted Local File Read Through Gate Artifact Import
Content: View full analysis
Remediation: View remediation

SkillSpector

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration

Findings (3)

Tp4

High

Category: MCP Tool Poisoning
Confidence: 95% confidence
Finding: The skill is presented as a strict human-gated planning workflow, but the documented behavior includes creating persistent session state and relying on conventions for approvals and deconfliction rather than enforceable controls. If users or operators trust the description, they may assume safety properties that are not actually guaranteed, allowing unauthorized writes, stale-session confusion, or bypass of the claimed human-review process.
Content: No source excerpt is available for this finding.

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 91% confidence
Finding: The README says users can invoke the skill by asking to 'start AIDLC for non-trivial work,' which is an ambiguous activation condition in markdown. It does not define what counts as 'non-trivial work' or provide exclusion examples, increasing the chance of unintended invocation.
Content: No source excerpt is available for this finding.

Undeclared Tool Scope

Medium

Category: MCP Least Privilege
Confidence: 85% confidence
Finding: The skill clearly instructs the agent to read and write workspace files and invoke local scripts, but it does not declare an explicit tool scope such as permissions or allowed-tools. That creates a trust and containment gap: a host may permit broader tool use than users expect, and the skill can perform filesystem mutations without a machine-readable restriction boundary.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.