Back to skill

Security audit

Pill Reminder

Security checks for vulnerabilities and agentic risk

Overview

This medication reminder skill is purpose-aligned, but it handles sensitive medication data in chat channels without enough privacy controls or scoping.

Review this skill carefully before installing. Use only private or tightly controlled messaging channels, avoid broad group chats, and assume medication names, refill counts, profile names, and adherence replies may be visible to the chat provider and channel members. The publisher should add private-channel defaults, confirmations, redaction, and authorization checks for config and log display.

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
SKILL.md:55
Finding

Full Medication Configuration Can Be Exposed Through Chat

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:55
Vulnerability Type: Sensitive information exposure through an unrestricted configuration-display command
Risk Level: Medium

Vulnerable Code Snippet

markdown
| `show config` | Display the current pills.md content. |

The sensitivity and scope of the exposed configuration are established in SKILL.md:138-139:

markdown
| Config | `~/.openclaw/workspace/pills.md` | All profiles, pills, schedules, counts |
| Log | `~/.openclaw/workspace/pill-log.md` | Adherence history, dates taken/missed |

The same command is also documented in README.md:88:

markdown
| `show config` | Display your current pills.md file. |

Technical Analysis

The show config command directs the agent to return the complete contents of pills.md. According to the documented data model, this file can contain profile names, medication names, dose schedules, remaining pill counts, refill thresholds, and information about multiple household members.

The skill operates through configurable messaging channels, including potentially shared GroupMe, Discord, or Telegram conversations. No documented control requires this command to run in a private channel, verifies that the requester is the configuration owner, limits the response to the requester's profile, requests confirmation, or redacts sensitive fields.

Consequently, raw health-related configuration data may be disclosed to unintended channel participants. This is an insecure handling and presentation practice for sensitive local data.

Attack Path

  1. A user configures the skill with one or more profiles and medication schedules.
  2. The skill stores those details in ~/.openclaw/workspace/pills.md.
  3. The bot is used in a shared or otherwise insufficiently private messaging channel.
  4. An unauthorized or unintended participant invokes show config, or the legitimate owner invokes it without realizing that the response will be visible to o ...[truncated 961 chars]
Remediation
View remediation

Remediation Suggestions

  1. Remove raw configuration display functionality from chat-facing commands, or replace it with a safe, structured summary.
  2. Require authenticated owner authorization before returning any configuration information.
  3. Refuse to display sensitive configuration in group channels; require a verified private conversation.
  4. Restrict output to the requester's own profile instead of returning all household profiles.
  5. Redact medication names, schedules, counts, and family-member identifiers unless each field is explicitly requested and disclosure is confirmed.
  6. Add a confirmation prompt that clearly identifies the destination channel and categories of data that will be exposed.
  7. Apply least-privilege access controls so family profiles can be viewed only by specifically authorized users.
  8. Avoid returning raw file contents. Parse the file and emit only allowlisted fields.
  9. Add security tests covering unauthorized users, shared channels, cross-profile access, redaction, and accidental group-chat disclosure.
  10. Update README.md and SKILL.md so their privacy claims clearly explain that chat output leaves local storage and may be retained by the messaging provider.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The README explicitly promotes sending medication reminders and adherence-related information through third-party chat platforms such as GroupMe, Discord, and Telegram, which can expose sensitive health information to external providers and potentially other channel members. Although the document mentions that data stays local unless sent via chat, it does not clearly foreground the privacy and confidentiality risks of transmitting medical information over those channels during setup or feature descriptions.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill sends reminder content containing profile names, pill names, and remaining counts to third-party messaging channels, but the description does not clearly warn users that sensitive medication data leaves the local workspace. This creates a privacy risk because medication/adherence information may be exposed through channel providers, shared chats, or misconfigured destinations.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The command model relies on very broad natural-language parsing ('Speak naturally') without clear trigger boundaries or confirmation requirements. In a messaging-based skill that handles cron creation, profile changes, and medication logging, this can cause accidental activation or misinterpretation of casual chat as commands, leading to unintended schedule changes or false adherence records.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

This template explicitly collects sensitive health-related information such as medication names, dosing schedules, refill thresholds, and profile names, and it pairs that data with a messaging channel setting. Even though it is only a template, the absence of any privacy warning, data-handling guidance, or caution against sharing through insecure group channels increases the risk of exposing medical and personal information.

Content

No source excerpt is available for this finding.