Security audit
Security checks for vulnerabilities and agentic risk
This decision-support skill is mostly a local Markdown workflow, but users should notice that it stores decision history locally and contains inconsistent callback/API-key wording.
Install only if you are comfortable with the agent creating a local ./decision-making folder that may contain sensitive decision context and preferences. Treat the callback_url and API key sections as inconsistent documentation; avoid using a callback URL unless the publisher clarifies what is sent and where, and do not store secrets in the skill files.
The manifest advertises purely local, no-external-system operation, but the interface includes external callback capability. Security-sensitive users may rely on the manifest when deciding whether the skill is safe to run, so this discrepancy undermines informed consent and can hide potential data egress paths.
The skill description emphasizes decision support but does not prominently warn that it stores decision history and preference signals locally. Users may disclose sensitive business, technical, or personal decision context without realizing it will be persisted on disk, creating confidentiality and retention risks.
The skill claims it never accesses external systems, yet its documented input includes a callback_url for outbound notification. This inconsistency can mislead users and reviewers about network behavior and data flow, increasing the risk of unintended external transmission of user decision data or metadata.
Using common phrases like 'that decision was right/wrong' as automatic review triggers can cause unintended invocation during ordinary conversation. In this skill, that means potentially writing retrospective records or updating learned preferences without clear user intent.
The trigger condition for running structured decision analysis is broad enough that normal conversations about tradeoffs may invoke the skill unexpectedly. Because the skill stores preferences and decision history locally and has write/exec capabilities, accidental activation can lead to unwanted persistence of sensitive user context.
The documentation states the skill requires no credentials, but later requires an API key and shows how to export it. Misrepresenting credential requirements can cause unsafe deployment assumptions, accidental secret exposure, or use in environments where operators believe no sensitive configuration is needed.
No suspicious patterns detected.
