Self-Criticism

• Outcome: Pass
• Latest audit: Sep 11, 2026 · 10:18 PM UTC
• Version: 1.0.0

Overview

This skill is a disclosed local workflow aid for adding self-critique checkpoints and keeping small local notes, with no evidence of exfiltration, hidden execution, or privilege escalation.

Install this if you want the agent to keep local self-critique preferences across sessions. Review what is written under ~/self-criticism/ and only approve SOUL or AGENTS edits if you want this workflow steering added to a workspace.

A.I.G

Vulnerability Patterns

• Skill Instruction Hijacking: Alters the agent's session goals or safety constraints when the skill loads
• Agent Memory Poisoning: Writes attacker-controlled rules into memory that affect later sessions
• Remote Payload Retrieval and Execution: Fetches external code whose behavior can change after review
• Embedded Malicious Code: Ships malicious scripts inside the skill and executes them locally
• Unauthorized Access and Privilege Escalation: Obtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns

• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (6)

Ae1

High

• Category: analysis-evasion

Content

- modifies its own `SKILL.md`

• Confidence: 100% confidence
• Finding: Referenced artifact was not completely inspected

Vague Triggers

Medium

• Confidence: 93% confidence
• Finding: The activation condition 'when another skill asks for a critique layer' and use when the directory is 'missing or empty' are broad enough that this skill may trigger in situations the user did not clearly intend. Because the skill also performs durable memory writes and workflow steering, ambiguous invocation increases the chance of unanticipated state changes and persistence.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill directs saving preferences and lessons to both main memory and local files without an explicit user-facing notice that durable storage will occur. This can lead to silent retention of behavioral preferences, friction history, or workflow lessons that the user may not expect to persist across sessions.

Sudo/Root Execution

Medium

• Category: Privilege Escalation

Content

```bash
mkdir -p ~/self-criticism/archive
touch ~/self-criticism/{memory.md,checkpoints.md,incidents.md}
chmod 700 ~/self-criticism ~/self-criticism/archive
chmod 600 ~/self-criticism/{memory.md,checkpoints.md,incidents.md}
```

• Confidence: 80% confidence
• Finding: Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Sudo/Root Execution

Medium

• Category: Privilege Escalation

Content

mkdir -p ~/self-criticism/archive
touch ~/self-criticism/{memory.md,checkpoints.md,incidents.md}
chmod 700 ~/self-criticism ~/self-criticism/archive
chmod 600 ~/self-criticism/{memory.md,checkpoints.md,incidents.md}
```

If `~/self-criticism/memory.md` is empty, initialize it from `memory-template.md`.

• Confidence: 80% confidence
• Finding: Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Vague Triggers

Medium

• Confidence: 93% confidence
• Finding: The trigger families are intentionally broad and lack concrete thresholds, exclusions, or prioritization rules, which can cause inconsistent activation across agents and contexts. In a safety-relevant skill, this ambiguity can lead to critique firing too often and being ignored, or not firing when high-risk transitions occur, reducing the reliability of the safeguard.

Static analysis

No suspicious patterns detected.