Back to skill

Security audit

root

Security checks for vulnerabilities and agentic risk

Overview

This is a local memory skill, but it persistently changes agent behavior and can activate an unpinned companion skill, so users should review it before installing.

Install only if you want the agent to save corrections and preferences locally and let those memories influence future sessions. Review the proposed AGENTS.md, SOUL.md, and HEARTBEAT.md edits before applying them, prefer Passive or Strict mode, periodically inspect or delete ~/self-improving, and decline the proactivity companion unless you trust the resolved package and version.

Vulnerability Patterns
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • 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 (2)

T02 · Agent Memory Poisoning

Warning
Location
setup.md:75
Finding

Persistent Modification of Core Agent Steering Files

Content
View full analysis
.md` - Do not read unrelated domains "just in case" If inferring a new rule, keep it tentative until human validation. ... - Explicit user correction → append to `~/self-improving/corrections.md` immediately - Reusable global rule or preference → append to `~/self-improving/memory.md` - Domain-specific lesson → append to `~/self-improving/domains/.md` - Project-only override → append to `~/self-improving/projects/.md` - Keep entries short, concrete, and one lesson per bullet; if scope is ambiguous, default to domain ra ...[truncated 3163 chars]
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
setup.md:89
Finding

Unpinned External Skill Installation and Immediate Activation

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (11)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill instructs the agent to automatically persist user corrections and preferences to local files, but it does not require an explicit just-in-time notice or consent at the moment memory is written. This creates a privacy and trust risk because users may not realize their statements are being stored durably, including information that later becomes sensitive in context.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The automatic logging rules capture natural-language user statements like preferences, corrections, and repeated instructions into persistent plain-text memory. Even with a stated boundary against credentials and health data, the collection trigger is broad enough that sensitive personal, project, or business information could be stored or later surfaced without adequate minimization or classification.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill exposes broad natural-language commands to search, list, and export accumulated memory, which can disclose large amounts of previously stored user data in one step. If an unauthorized user, prompt injection, or mistaken request reaches the agent, these retrieval and export paths increase the blast radius of whatever sensitive material was retained.

Content

No source excerpt is available for this finding.

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 178)May include surrounding context.

md
- Pattern used 3x in 7 days → promote to HOT
- Pattern unused 30 days → demote to WARM
- Pattern unused 90 days → archive to COLD
- Never delete without asking

### 4. Namespace Isolation
- Project patterns stay in `projects/{name}.md`

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 · boundaries.md (reported line 11)May include surrounding context.

md
| Financial | Card numbers, bank accounts, crypto seeds | Fraud risk |
| Medical | Diagnoses, medications, conditions | Privacy, HIPAA |
| Biometric | Voice patterns, behavioral fingerprints | Identity theft |
| Third parties | Info about other people | No consent obtained |
| Location patterns | Home/work addresses, routines | Physical safety |
| Access patterns | What systems user has access to | Privilege escalation |

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The phrase "forget everything" is a broad natural-language trigger that could be invoked accidentally during ordinary conversation, quotations, testing, or discussion about the feature itself. In a self-improving, memory-bearing agent, unintended activation could cause destructive deletion of learned state and service disruption, especially if the phrase is acted on without confirmation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill specifies automatic loading of memory and project context at session start without any explicit notice, consent gate, or visibility control. This can cause stored personal preferences, prior conversation data, or project-specific context to be accessed unexpectedly, increasing privacy risk and making silent cross-session data use more likely.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The correction-handling flow automatically persists user inputs, timestamps, counters, and context to disk with no user-facing warning or consent step. This creates a risk that sensitive data, one-off instructions, or unintended personal information will be retained permanently and later reused outside the user's expectations.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The weekly maintenance routine performs automated scanning, movement, archiving, and compaction of memory files without informing the user about these background data-management actions. Even if intended for housekeeping, such silent processing can expand retention scope, duplicate sensitive content into archives, and reduce the user's ability to understand where their data resides.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The guidance explicitly recommends "aggressive learning" with "minimal confirmation" for power users, which can cause the agent to persist inferred preferences or behavioral changes without clear user approval. In a self-improving memory skill, that increases the chance of unwanted memory writes, silent preference drift, and incorrect cross-task personalization that affects future outputs.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

The template explicitly instructs creation of directories and files in the user's home directory without any warning, confirmation step, or scope limitation. While the commands are simple and not inherently destructive, they normalize persistent filesystem modification by the skill and could lead to unexpected writes, privacy issues, or unsafe automation when executed by an agent without clear user consent.

Content

No source excerpt is available for this finding.