Back to skill

Security audit

lean-startup

Security checks for vulnerabilities and agentic risk

Overview

This skill provides a coherent startup experiment framework, and the flagged phrases describe market uncertainty rather than suppressing safety warnings.

Use it to plan experiments. Separately authorize customer-facing tests or spending, and keep confidential customer details out of reusable observations.

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
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

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 · examples/ai-native-lean-startups-2023-2026.md (reported line 9)May include surrounding context.

md
feature, raises on the demo, and then a subsequent model release from the underlying provider (or an open-weight model) absorbs that feature into the base capability — the "GPT-wrapper gets wrapped" problem. The dramatic public reminder came in **January 2025**, when the Chinese lab **DeepSeek** released a strong, low-cost open-weight reasoning model; the reaction rippled through markets and, on **27 January 2025**, Nvidia's share price fell sharply in a single session — a widely reported signal that the cost and moat assumptions underpinning many AI plans could shift without warning.

The Lean Startup correction: in an AI-native startup, the load-bearing assumption is almost never "can we build the feature?" (you can — cheaply). It is **"does a specific customer keep using and paying for the workflow *after* the underlying model capability becomes a commodity available to everyone?"** That is a retention-and-willingness-to-pay assumption, and it is exactly what a demo does not test.

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 · references/sources.md (reported line 11)May include surrounding context.

md
v=7QmCUDHpNzE
- **Sequoia Capital, Dropbox pitch deck, 2007** — primary-source artifact from the early Dropbox story.
- **OpenAI.** "Introducing ChatGPT." 30 November 2022. https://openai.com/index/chatgpt/ — public marker for the foundation-model API wave that made the *Build* phase nearly free for AI-native startups (2023–2026 example).
- **Reuters / Associated Press / Financial Times**, late January 2025 coverage of DeepSeek's low-cost open-weight model release and the ~27 January 2025 selloff in Nvidia and AI-linked equities — durable, widely-reported reminder that AI moat/cost assumptions can reset without warning (2023–2026 example). Exact intraday figures varied by source and are cited qualitatively.
- The popular framing "fail fast" is **not** cited here as a source — by this skill's own rule, an aphorism is not evidence. Lean Startup is more precisely "**test cheaply** the demand-side assumption *before* paying to build the supply-side capability"; failure speed is incidental.