Security audit
Security checks for vulnerabilities and agentic risk
This is a startup-diagnostic skill that asks questions, does purpose-aligned research, and saves a report, with no evidence of hidden code, persistence, or exfiltration.
Install only if you are comfortable with the agent using web search, reading product-related repository files and recent git history, and writing a diagnostic report under docs/business. Avoid using it in repositories that contain unrelated confidential plans, credentials, or sensitive customer data unless you first limit what the agent can read.
Skill manipulates agent memory, state, or stored context. Memory corruption can alter personality, override safety rules, or cause unpredictable behavior.
Output premises as clear statements:
PREMISES:
1. [statement] — agree/disagree?
The trigger phrases are very broad, covering generic requests like 'brainstorm this' or 'I have an idea.' That increases the chance the skill activates in situations where the user did not ask for an aggressive diagnostic workflow with web search, file inspection, and report writing.
The skill says it 'produces a diagnostic report, not code,' but then instructs the agent to perform external web searches and inspect the local repository. That creates a capability mismatch: users may reasonably expect a conversational startup diagnostic, while the skill silently expands scope into network access and local data access.
The skill directs the agent to save a report to docs/business/office-hours-report.md without warning the user or obtaining consent. Silent file creation/modification can surprise users, overwrite existing content, or persist sensitive business analysis in a repository unintentionally.
Reading README, AGENTS.md, product files, and recent git history is broader than necessary for a generic startup-idea diagnostic. In many environments, repository files and commit history can contain sensitive internal plans, credentials, customer names, or unrelated project context that the user did not intend to expose for this task.
No suspicious patterns detected.