Security audit

Section 11: Endurance Training Coach (Intervals.icu)

Security checks for vulnerabilities and agentic risk

Overview

This endurance-coaching skill is mostly coherent, but it asks agents to trust mutable remote protocol files and to modify persistent agent instructions, which users should review before installing.

Install only if you are comfortable with the agent reading sensitive training and wellness data and with the optional Intervals.icu write workflow. Prefer local packaged protocol/templates or pinned GitHub revisions, keep GitHub data repositories private, avoid adding this skill's rule to global agent instructions unless you understand the cross-session effect, and require explicit confirmation before any calendar or threshold changes.

A.I.G

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 (2)

T01 · Skill Instruction Hijacking

Warning

Location: SKILL.md:49
Finding: Mutable Remote Instructions Are Loaded as Authoritative Coaching Protocol
Content: View full analysis
Remediation: View remediation

T02 · Agent Memory Poisoning

Warning

Location: SKILL.md:41
Finding: Skill Instructs Modification of Global Persistent Agent Configuration
Content: View full analysis
Remediation: View remediation

SkillSpector

Vulnerability Patterns

MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
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 (2)

Intent-Code Divergence

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The skill’s privacy/security section states that fetched content comes only from user-configured sources, but the protocol explicitly falls back to hardcoded public GitHub raw URLs for templates and protocol files. This creates a trust-boundary mismatch: if local/repo files are absent, the agent may silently ingest remote unpinned content from a third party, enabling supply-chain tampering or unexpected data flow contrary to user expectations.
Content: No source excerpt is available for this finding.

Missing User Warnings

Low

Category: Not specified by scanner
Confidence: 84% confidence
Finding: The template instructs the agent to write summaries and plans to a user-chosen location but does not warn about sensitive training/health data exposure, accidental overwrites, or unsafe destinations. In this coaching context, outputs may contain personal wellness and performance information, so ambiguous write targets can lead to unintended disclosure or data loss even without malicious intent.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.