Master Security Toolkit · Prompt library

Ask for something useful.

Choose a task, copy its prompt and add your own material. Bracketed fields are yours to replace.

Check before installing

Bring: The source and intended use.
Expect: A verdict with evidence.

Is [package or repo] safe to install for [purpose]? Read its scripts and manifests and give me a verdict, confidence and safer path.

Bring: The MCP's configuration.
Expect: A capability fingerprint and verdict.

Review this MCP before I connect it. Fingerprint its launch command, tools, permissions and auth, and name the risks.

Bring: A local package.
Expect: Line-level findings.

Scan this local skill package for risky instructions, broad permissions, downloads and secrets. Confirm each finding at the line level.

Bring: Prompts, code or config.
Expect: Findings and tests.

Assess this AI feature for prompt-injection risk across its prompts, retrieval and tools. Recommend fixes and regression tests.

Bring: Sample retrieved content.
Expect: Entry points and mitigations.

Our assistant reads [source]. Review it for hidden instructions and what they could reach, with a safe test I can run.

Bring: The artifact and its claimed source.
Expect: A provenance verdict.

Verify this download against its claimed release: checksum, signature and source traceability. Tell me what's missing.

Bring: The manifest or plist.
Expect: A least-privilege review.

This [app or script] asks for [permissions]. Check each request against its purpose and recommend the least privilege that works.

Bring: A confirmed finding.
Expect: Mitigations and fixtures.

Turn this confirmed finding into layered mitigations and pass/fail regression tests. State the residual risk.