better-readme

• Outcome: Pass
• Latest audit: Jul 31, 2026 · 6:21 PM UTC
• Version: 1.1.1

Overview

This README-assistant skill is purpose-aligned and disclosed, with only opt-in public-repository review and local project scanning to consider before use.

Installers should expect the skill to read README and project metadata, run its local Python scoring script, and possibly offer an opt-in review of public GitHub repository READMEs. Decline the onboarding repo audit if you want local-only behavior, and confirm language/file targets before allowing it to generate or overwrite README files.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration

Findings (16)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 94% confidence
• Finding: The skill instructs the agent to read repository files and execute a shell command (`python3 scripts/readme_audit.py`) but does not declare permissions for file access or shell execution. This creates a transparency and policy-enforcement gap: users and the runtime may not have clear notice that the skill can inspect files and run local code, which increases the chance of overbroad access or unsafe execution.

Description-Behavior Mismatch

Medium

• Confidence: 95% confidence
• Finding: The README expands the skill's behavior from README assistance into fetching and batch-auditing the user's GitHub repositories during onboarding. That creates an unnecessary data-access path and scope creep for a documentation skill, increasing privacy risk and the chance an agent performs repository enumeration the user did not explicitly request.

Context-Inappropriate Capability

Medium

• Confidence: 97% confidence
• Finding: Claiming the skill can fetch a user's GitHub repositories is unjustified for the stated purpose of improving README files. Even if intended as convenience, it could prompt broad access to source code and metadata unrelated to the immediate task, creating avoidable privacy and over-collection concerns.

Description-Behavior Mismatch

Medium

• Confidence: 91% confidence
• Finding: The onboarding flow expands the skill from README authoring into broad repository enumeration and batch evaluation of the user's public GitHub repos. Even if limited to public repos, this is a scope jump that can surprise users, trigger unnecessary network access, and collect more contextual information than needed for the immediate task.

Context-Inappropriate Capability

Medium

• Confidence: 95% confidence
• Finding: Fetching a user's public GitHub repository list is not necessary for the core task of improving a README and introduces avoidable network-based data collection. In this skill context, the capability is more dangerous because the user likely expects document help, not account-level repository discovery, which can erode consent and privacy boundaries.

Vague Triggers

Medium

• Confidence: 82% confidence
• Finding: The trigger phrase 'Check my README' is broad and conversational, which increases the chance of accidental activation in normal discussion. In the context of a skill that may scan projects or initiate onboarding actions, ambiguous triggering can lead to unintended file analysis or follow-on prompts for wider access.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The README describes first-use fetching of GitHub repositories without a prominent privacy warning or clear explanation of data-access implications. Users may not expect a README skill to enumerate repositories, so insufficient disclosure increases the risk of uninformed consent and unnecessary exposure of project metadata or code.

Vague Triggers

Medium

• Confidence: 84% confidence
• Finding: The trigger examples in Agent Integration are broad enough to overlap with ordinary user requests, which can cause the skill to activate in situations where the user did not intend expanded workflow behavior. Given this README also advertises scanning and onboarding actions, weak trigger boundaries increase the chance of unintended analysis or escalation to privacy-sensitive steps.

Vague Triggers

Medium

• Confidence: 89% confidence
• Finding: The trigger phrase encourages activation from a very common natural-language request without clear scoping or exclusion criteria. In an agent ecosystem, broad triggers can cause the skill to activate in contexts the user did not intend, leading to unnecessary repository scanning or README modification workflows.

Vague Triggers

Medium

• Confidence: 92% confidence
• Finding: The listed example phrases are generic requests that overlap with ordinary conversation and do not define when the skill should or should not run. This increases the chance of accidental invocation and downstream actions like codebase inspection or file generation without sufficiently specific user intent.

Natural-Language Policy Violations

Medium

• Confidence: 84% confidence
• Finding: The documentation promises automatic bilingual output as a fixed behavior rather than as a user-selected option. Forcing output language choices can override user intent, create unnecessary file changes, and in some environments lead to unintended publication of extra-language content.

Natural-Language Policy Violations

Medium

• Confidence: 90% confidence
• Finding: The workflow hard-codes generation of both README.md and README.zh-CN.md without any consent or preference check. In a documentation skill, that means the agent may create or overwrite files beyond the user's requested scope, which is especially risky in automated or CI-assisted contexts.

Natural-Language Policy Violations

Medium

• Confidence: 88% confidence
• Finding: Describing the agent as always generating a bilingual README removes user choice and can expand the action surface beyond the requested task. In this skill context, the danger is not code execution but unintended content generation and modification of repository documentation artifacts.

Vague Triggers

Medium

• Confidence: 87% confidence
• Finding: Broad activation phrases like 'my README is bad' or 'I need a README template' can cause the skill to trigger in loosely related conversations, increasing the chance that it reads files or initiates workflows unexpectedly. Because this skill can inspect the codebase and run an audit script, accidental invocation carries more risk than a purely advisory skill.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The skill proposes fetching the user's public GitHub repo list without an upfront privacy or network-access warning. In a documentation-focused skill, hidden or underexplained network activity is risky because users may not anticipate external lookups or understand the extent of account-level enumeration.

Vague Triggers

Medium

• Confidence: 90% confidence
• Finding: The Agent Skill README template encourages generic trigger phrases and broad context matching without requiring narrowing conditions, exclusions, or safety boundaries. In an agent ecosystem, ambiguous triggers can cause a skill to activate unexpectedly for unrelated prompts, increasing the chance of misrouting user intent, conflicting with other skills, or running documentation workflows in inappropriate contexts.

VirusTotal

65/65 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.