ClawHub
Back to skill

Security audit

stoic-coach

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed local Stoic journaling coach, with sensitive but purpose-aligned local storage and no evidence of hidden network, destructive, or deceptive behavior.

Install only if you are comfortable with plaintext local journaling. Avoid entering secrets or highly sensitive details unless the storage directory is protected, be careful when using a synced data directory or exporting, and pause heavier mortality/loss exercises if they feel destabilizing.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • 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
Findings (4)

Tainted flow: 'JOURNAL_FILE' from os.environ.get (line 34, credential/environment) → open (file write)

Medium
Category
Data Flow
Content
def save_entry(entry: dict) -> dict:
    os.makedirs(DATA_DIR, exist_ok=True)
    with open(JOURNAL_FILE, "a", encoding="utf-8") as f:
        f.write(json.dumps(entry, ensure_ascii=False) + "\n")
    return entry
Confidence
90% confidence
Finding
The script trusts STOIC_COACH_DATA_DIR from the environment and uses it to construct JOURNAL_FILE, which is then opened for append without validation. In a host/agent environment where environment variables can be influenced, this can redirect sensitive journaling data to an unintended path or overwrite/append to attacker-chosen files accessible to the process.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
These sections introduce mortality, loss, and emotionally intense reflection exercises that can destabilize vulnerable users, yet the guidance lacks a clear up-front warning, screening step, or explicit stop/grounding escalation before entering the exercise. In a mental-wellness coaching skill aimed at anxious, ruminative, and self-doubting users, this omission increases the chance of worsening distress, triggering grief spirals, or encouraging the model to persist in deep emotional content when the user is not well resourced.

Missing User Warnings

Medium
Confidence
83% confidence
Finding
The tool stores intimate reflection data, mood/state scores, tags, and free-text insights locally and supports export, but the user-facing header does not warn that this sensitive data is retained on disk and may be reproduced later. For a mental-health-adjacent journaling skill, this omission materially increases privacy risk because users may disclose highly sensitive content without informed consent about storage and export behavior.

Ssd 3

Medium
Confidence
92% confidence
Finding
The script intentionally retains and later reprints users' free-text reflections and insights in history, insight reports, map views, and exports. In this context, the data is likely to include sensitive mental-health, work, family, or identity details, so any downstream display, logging, sharing, or export can leak highly revealing natural-language content beyond the user's original expectation.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.