Back to skill

Security audit

Adhd Assistant Repo

Security checks for vulnerabilities and agentic risk

Overview

This ADHD productivity skill is not malicious, but it merits Review because it could persist sensitive health and treatment details and activate on ordinary productivity requests without explicit consent or retention limits.

Install only if you are comfortable with an ADHD-framed assistant storing personal productivity and health-adjacent patterns. Before using memory features, ask the agent not to save diagnosis, medication, therapy, triggers, or sensitivities unless you explicitly approve each memory, and prefer general preferences such as reminder style or focus times.

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
Findings (1)

other

Warning
Location
SKILL.md:128
Finding

Persistent Storage of Sensitive Health and Behavioral Information Without Data-Minimization Controls

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:128-143 and SKILL.md:365-370
Vulnerability Type: Excessive Sensitive Health Data Collection and Retention
Risk Level: Medium

Vulnerable code:

markdown
## User Preferences to Learn

Over time, remember these preferences (via SkillBoss API Hub memory):

**Schedule & Energy:**
- Peak focus hours (morning person vs. night owl)
- Typical energy patterns throughout the day
- Best times for deep work vs. shallow tasks

**Task Management:**
- Preferred number of daily priorities (1-3 recommended)
- Task/note storage location (files, apps, directories)
- Preferred reminder frequency and channels

**ADHD Profile:**
- Diagnosed or suspected ADHD
- Current treatments (medication, therapy) - for context only
- Common pitfalls (social media, hyperfocus traps)
- Strategies that have worked in the past
markdown
### Memory Usage:
- Store user preferences and patterns
- Remember what strategies have worked
- Track routine adherence over time
- Note energy patterns and triggers

Technical Analysis

The skill instructs the agent to place potentially sensitive health and behavioral information into the SkillBoss API Hub's persistent memory. The listed information includes diagnosed or suspected ADHD, medication or therapy, routine adherence, energy patterns, and emotional or behavioral triggers.

This exceeds data-minimization principles because treatment information is explicitly retained “for context only,” rather than being necessary to perform ordinary planning and task-management functions. The instructions do not require explicit informed consent before persistence, restrict which medical details may be stored, define a retention period, provide deletion procedures, or require redaction before memory writes.

This is not classified as T02 Agent Memory Poisoning because the retained content is user profile data rather than ...[truncated 1802 chars]

Remediation
View remediation

Remediation Suggestions

  1. Do not persist diagnosis status, medication details, therapy details, or detailed emotional triggers by default.
  2. Require explicit, informed, and granular opt-in consent before storing any health-related information.
  3. Limit persistent memory to task-relevant operational preferences, such as preferred reminder times or checklist format.
  4. Store generalized preferences instead of source medical details. For example, retain “prefers short reminders” rather than the diagnosis or treatment context used to derive that preference.
  5. Add a pre-write review that tells the user exactly what information will be stored and allows modification or rejection.
  6. Define retention periods and automatically expire sensitive information when it is no longer required.
  7. Provide accessible commands to inspect, correct, export, and delete all stored memories.
  8. Redact health details from task files, calendar entries, review summaries, and integrations unless the user separately authorizes their inclusion.
  9. Restrict downstream task-tracker, calendar, and note-taking integrations from receiving health information by default.
  10. Document the memory system's access controls, encryption expectations, retention behavior, and incident-response process.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (5)

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
80% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · SKILL.md (reported line 49)May include surrounding context.

md
- Provides virtual body doubling sessions
- Creates structured co-working check-ins
- Sets up accountability partnerships
- Offers presence-based support without judgment

### 6. Dopamine Regulation
- Helps build personalized "dopamine menus"

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The top-level description says to use the skill for broad scenarios like organizing life, planning the day/week, beating procrastination, or managing overwhelm. Those conditions overlap heavily with ordinary productivity requests, so the skill may activate for users who did not ask for ADHD-specific support and may then steer conversations toward mental-health framing unnecessarily.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The activation list includes common states like feeling overwhelmed, needing help planning, procrastination, forgetfulness, and wanting productivity help. Without clear exclusions, the router could apply this mental-health-adjacent skill to many ordinary requests, increasing the chance of inappropriate personalization, sensitive data collection, or misleading ADHD framing.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger phrases are everyday statements such as 'I need help planning' and 'I feel overwhelmed,' which are too generic for reliable routing. This creates a realistic risk of over-activation, causing users to receive ADHD-specific behavioral guidance and memory collection when they only wanted general assistance.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill explicitly proposes remembering diagnosis status, suspected ADHD, treatment context, sensitivities, and behavioral patterns via memory. That is sensitive mental-health information, and the file does not require clear disclosure, consent, minimization, retention limits, or an opt-out before persistence, which creates privacy and compliance risk.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.