Back to skill

Security audit

reminder-engine

Security checks for vulnerabilities and agentic risk

Overview

This reminder skill is not clearly malicious, but it asks for broad command/file authority and can schedule external notifications with under-scoped guidance.

Review this skill before installing. Use it only for explicit reminder tasks, avoid sending sensitive conversation content through Discord, Telegram, WhatsApp, or webhooks, replace all sample secrets with strong generated values, and regularly list or clean up scheduled jobs created by the skill.

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:124
Finding

Predictable Hardcoded Webhook Signing Secrets in Executable Examples

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:124-131, SKILL.md:240-247, and SKILL.md:252-259
Vulnerability Type: Hardcoded and predictable webhook signing secrets
Risk Level: Medium

Vulnerable Code

SKILL.md:124-131:

bash
skill-platform cron add \
  --name "reminder-deploy-verify" \
  --at "$REMIND_AT" \
  --session isolated \
  --message "验证最新部署的功能" \
  --webhook "https://monitoring.example.com/hooks/verify" \
  --webhook-signing-secret "monitor-secret" \
  --delete-after-run

SKILL.md:240-247:

bash
skill-platform cron add \
  --name "reminder-deploy-verify" \
  --at "$(date -u -d '+5 minutes' +"%Y-%m-%dT%H:%M:%SZ")" \
  --session isolated \
  --message "验证最新部署的功能" \
  --webhook "https://ci.example.com/hooks/verify" \
  --webhook-signing-secret "ci-secret" \
  --delete-after-run

SKILL.md:252-259:

bash
skill-platform cron add \
  --name "reminder-incident-alert" \
  --at "$(date -u -d '+1 minute' +"%Y-%m-%dT%H:%M:%SZ")" \
  --session isolated \
  --message "触发告警:生产事故" \
  --webhook "https://alerts.example.com/hooks/incident" \
  --webhook-signing-secret "alert-secret" \
  --delete-after-run

Technical Analysis

The documentation provides copy-paste-ready commands containing static, predictable webhook signing secrets such as monitor-secret, ci-secret, and alert-secret. Although these appear in examples rather than as deployed production credentials, users may execute the examples after changing only the endpoint. This can result in the documented value becoming an actual authentication secret.

A webhook signature only proves authenticity while its signing key remains secret and sufficiently unpredictable. An attacker who knows or guesses one of these values can construct an arbitrary request body and calculate the expected HMAC signature. If the receiving endpoint relies solely on that signature, it cannot distingui ...[truncated 2248 chars]

Remediation
View remediation

Remediation Suggestions

  1. Replace all literal signing secrets with references to protected runtime values:

    bash
    : "${WEBHOOK_SIGNING_SECRET:?WEBHOOK_SIGNING_SECRET must be configured}"
    
    skill-platform cron add \
      --name "reminder-deploy-verify" \
      --at "$REMIND_AT" \
      --session isolated \
      --message "Verify the latest deployment" \
      --webhook "$WEBHOOK_URL" \
      --webhook-signing-secret "$WEBHOOK_SIGNING_SECRET" \
      --delete-after-run
    
  2. Generate an independent, cryptographically random secret for each integration, with at least 256 bits of entropy. Do not reuse one secret across monitoring, CI/CD, and incident-management receivers.

  3. Prefer a platform secret manager or protected credential reference over direct command-line arguments. If the CLI supports secret identifiers, file descriptors, or standard-input ingestion, use those mechanisms to prevent exposure through process listings and shell history.

  4. Ensure the command fails closed when the secret is absent. Never silently fall back to a sample or default value.

  5. Mark all example values explicitly as nonfunctional placeholders and warn users that they must not be deployed unchanged.

  6. On the receiving endpoint, use constant-time signature comparison, enforce HTTPS certificate validation, reject unsigned requests, validate timestamps and nonce values to prevent replay, and limit the maximum accepted request age.

  7. Rotate any signing key that was based on these examples and audit webhook logs for forged, replayed, or otherwise unexpected requests.

  8. Restrict each receiving integration to the minimum downstream privileges needed for reminder delivery so that a forged webhook cannot trigger unrelated administrative or deployment operations.

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Vague Triggers

High
Category
Not specified by scanner
Confidence
94% confidence
Finding

The activation guidance is excessively broad ('use when AI model calling, orchestration, LLM applications') and paired with powerful tools like exec/write plus webhook and multi-channel delivery. In an agent environment, vague activation criteria can cause the skill to be invoked outside its intended reminder scope, leading to unintended command execution, message dispatch, or external network actions with user data.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill prominently supports Discord, Telegram, WhatsApp, and webhook delivery but does not provide a clear up-front warning that user content may be transmitted to third-party services. In this context, the danger is elevated because the skill is designed for agent-driven automation and includes external delivery channels, so sensitive conversation content could be sent off-platform without sufficiently explicit user awareness or consent.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest description at L011-L012 defines the skill's purpose as enterprise reminder creation with advanced scheduling and delivery features. However, the capability table adds '深度漏洞扫描与CVE关联' and '安全基线合规审计', which are semantically separate security-audit functions and not an obvious implementation detail of reminder creation.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill is described throughout the manifest as creating and managing reminders, channels, schedules, and webhook deliveries. But the declared input parameters ('content', 'mode', 'style') and output schema ('处理结果', template/style metadata) describe a generic text-processing interface rather than reminder-specific operations, indicating a mismatch between the claimed reminder behavior and the documented interface.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

Line L344 states the skill '提供可视化界面,简化设置流程', which actively asserts a visual UI capability. Elsewhere the manifest consistently describes MD+EXEC usage through agent instructions and command-line examples, with no corresponding UI implementation or interface definition in the file.

Content

No source excerpt is available for this finding.