Back to skill

Security audit

Dual-System Thinking (System 1 / System 2)

Security checks for vulnerabilities and agentic risk

Overview

This skill provides a structured decision-making guide with no evidence of hidden access, unsafe execution, or data collection.

Use this as a thinking aid and independently verify consequential claims. Choose where decision logs are kept and avoid recording unnecessary sensitive details.

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
  • 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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (5)

Instruction Override

High
Category
Prompt Injection
Confidence
90% confidence
Finding

This pattern attempts to override system instructions or ignore safety constraints. Without LLM analysis, manual review is recommended.

Content

Scanner excerpt · examples/kahneman-2011-tversky-kahneman-research.md (reported line 17)May include surrounding context.

md
Over the following 30+ years, dozens of additional biases were documented: framing effects, sunk-cost reasoning, hindsight bias, confirmation bias, the planning fallacy, the conjunction fallacy, the base-rate fallacy, etc.

The dual-system framework was Kahneman's 2011 synthesis. It answered the question "why are humans biased in these systematic ways?" with: *because System 1 produces fast heuristic answers that work in familiar situations and fail in novel/statistical/high-stakes ones, and System 2 is too lazy to override System 1 most of the time.*

Kahneman's most operationally cited passage:

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · examples/kahneman-2011-tversky-kahneman-research.md (reported line 10)May include surrounding context.

md
The 1974 *Science* paper "Judgment under Uncertainty: Heuristics and Biases" articulated three foundational heuristics:

- **Representativeness** — judging probability by similarity to a stereotype (e.g., the "Linda problem" — subjects judge Linda more likely to be a "feminist bank teller" than a "bank teller," violating basic probability)
- **Availability** — judging frequency by ease of recall (see [`availability-heuristic`](../skills/availability-heuristic/SKILL.md))
- **Anchoring and adjustment** — being unduly influenced by an initial number (see [`anchoring`](../skills/anchoring/SKILL.md))

The 1979 *Econometrica* paper "Prospect Theory" articulated the value function and probability-weighting function that account for choice under risk (see [`loss-aversion-prospect-theory`](../skills/loss-aversion-prospect-theory/SKILL.md)).

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · examples/kahneman-2011-tversky-kahneman-research.md (reported line 11)May include surrounding context.

md
- **Representativeness** — judging probability by similarity to a stereotype (e.g., the "Linda problem" — subjects judge Linda more likely to be a "feminist bank teller" than a "bank teller," violating basic probability)
- **Availability** — judging frequency by ease of recall (see [`availability-heuristic`](../skills/availability-heuristic/SKILL.md))
- **Anchoring and adjustment** — being unduly influenced by an initial number (see [`anchoring`](../skills/anchoring/SKILL.md))

The 1979 *Econometrica* paper "Prospect Theory" articulated the value function and probability-weighting function that account for choice under risk (see [`loss-aversion-prospect-theory`](../skills/loss-aversion-prospect-theory/SKILL.md)).

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · examples/kahneman-2011-tversky-kahneman-research.md (reported line 13)May include surrounding context.

md
- **Availability** — judging frequency by ease of recall (see [`availability-heuristic`](../skills/availability-heuristic/SKILL.md))
- **Anchoring and adjustment** — being unduly influenced by an initial number (see [`anchoring`](../skills/anchoring/SKILL.md))

The 1979 *Econometrica* paper "Prospect Theory" articulated the value function and probability-weighting function that account for choice under risk (see [`loss-aversion-prospect-theory`](../skills/loss-aversion-prospect-theory/SKILL.md)).

Over the following 30+ years, dozens of additional biases were documented: framing effects, sunk-cost reasoning, hindsight bias, confirmation bias, the planning fallacy, the conjunction fallacy, the base-rate fallacy, etc.

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · examples/kahneman-2011-tversky-kahneman-research.md (reported line 41)May include surrounding context.

md
**Marketing and product design.** Dual-system insight underlies modern marketing — System 1 responses to packaging, branding, scarcity signals; System 2 engagement for high-consideration purchases. Conversion-optimization, dark patterns, and consent-flow design all operate on the framework.

**Public policy and choice architecture.** Sunstein and Thaler's "nudge" framework operationalizes the dual-system insight: design defaults and choice presentations such that System 1's tendencies produce socially-beneficial outcomes (e.g., opt-out 401k enrollment, organ donation defaults). See [`hyperbolic-discounting`](../skills/hyperbolic-discounting/SKILL.md).

**Executive decision-making.** The Kahneman-cited research on executive over-confidence, planning fallacy, and bias in M&A decisions has shaped how sophisticated decision-makers structure their own processes — premortems, devil's advocates, decision journals, all are explicit System 2 recruitment.