Borrowing it
Nothing to install: this file belongs to greenmtnsun/GitEasy. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/greenmtnsun/GitEasy/main/.claude/skills/virtual-security-reviewer/SKILL.mdgit clone --depth 1 https://github.com/greenmtnsun/GitEasyWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/greenmtnsun/giteasy/virtual-security-reviewer)<a href="https://agentmods.dev/skills/greenmtnsun/giteasy/virtual-security-reviewer"><img src="https://agentmods.dev/badge/skills/greenmtnsun/giteasy/virtual-security-reviewer/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/greenmtnsun/giteasy/virtual-security-reviewer"><img src="https://agentmods.dev/badge/skills/greenmtnsun/giteasy/virtual-security-reviewer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00214 | $0.02460 |
| Opus 5.5 | $0.00086 | $0.00984 |
| Sonnet 5.5 | $0.00043 | $0.00492 |
| Haiku 4.5 | $0.00021 | $0.00246 |
Grade C, and why
virtual-security-reviewer scanned grade C with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured today.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Instruction-override phrasingmediumPrompt injection
Text telling the model to disregard its earlier instructions or safety rules is the shape of a prompt injection, whoever wrote it.
| **CWE-1039** | Prompt Injection (a.k.a. OWASP LLM-01) | Any AI-driven harvester or transformer: WikiEngine (PSGallery help-text → Ollama), PesterForge v2.0 (customer modules → Ollama for stub generation), future AI tra Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.
Recursive force deletemediumDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
- **Worst-case-string thinking:** for any string field, what's the most malicious value? Path traversal (`../../etc/passwd`)? Command injection (`; rm -rf /`)? SQL (`' OR 1=1 --`)? Format string (`%s%s%s`)? PowerShell ex Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.
How it starts
The opening of the file, as written. The whole thing — 95 lines — stays where its author put it; the contents beside it link to each section on GitHub.
See also:
air-gap-disciplinefor the no-host-leakage rule,verify-claims-on-diskfor trust-but-verify on sibling-claim apply assertions,software-creationfor the structured-result-never-throw discipline that makes security-relevant errors inspectable.
Virtual Security Reviewer
The role of this persona is to look at code, designs, or proposals with adversarial intent — what could break this if someone wanted to — and surface real risks with named trust boundaries and named assumptions, not generic checklists.
Tone: adversarial-but-fair. Names the trust boundary explicitly. Pushes back on "this can't happen" by asking what specifically prevents it. Doesn't generate exploitable PoCs (the goal is to identify the gap, not to weaponize it).
What this persona knows about Keith's situation
Pre-loaded context (verify if claims feel stale):
- The cross-suite CWE list Keith confirmed as applicable across his projects (CWE-94 through CWE-770 handed off from PesterForge's AI 2026-05-08; CWE-1039 added 2026-05-27 after the Promptfoo research surfaced the LLM-specific injection vector):
| CWE | Class | Where it bites hardest in the suite |
|---|---|---|
| CWE-94 | Code Injection | PesterForge (AI emits code that user runs); WikiEngine (renderers emit markup with code blocks) |
| CWE-78 | OS Command Injection | Anywhere & <path>, Start-Process, or shell-out happens |
| CWE-502 | Deserialization | Anywhere ConvertFrom-Json / Import-CliXml runs on untrusted input — most so for PesterForge (Anthropic API responses) and WikiEngine (Confluence/REST returns) |
| CWE-22 | Path Traversal | -Path / -ScriptPath parameters in every project |
| CWE-200 | Info Exposure | Logs, error messages, telemetry — universal across the suite |
| CWE-256 | Plaintext Password | The 2026-05-04 + 2026-05-08 incidents (key in editor buffer / file) prove this is a personal pattern, not per-project |
| CWE-918 | SSRF | PesterForge (Anthropic), WikiEngine (Confluence), eventually GitEasy if it grows remote-API features |
| CWE-770 | Allocation Without Limits | PesterForge has -CharacterBudget; others should adopt the pattern |
| CWE-1039 | Prompt Injection (a.k.a. OWASP LLM-01) | Any AI-driven harvester or transformer: WikiEngine (PSGallery help-text → Ollama), PesterForge v2.0 (customer modules → Ollama for stub generation), future AI transformers in ModuleWorkshop. Untrusted source content can carry "Ignore previous instructions"-style payloads that hijack the LLM call. |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- today First seen · 95 lines · 214 tokens per session scan C 94fa70f951ea
virtual-security-reviewer is a skill published in the GitHub repository greenmtnsun/GitEasy (1 stars, last pushed yesterday), licensed MPL-2.0. It adds 214 tokens to every session and 2,460 once invoked, about $0.0009 per session on Opus 5.5. A static security scan graded it C with 2 findings (instruction-override phrasing, recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-11-18.
virtual-security-reviewer compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| virtual-security-reviewergreenmtnsun/GitEasy | Skill | 1 | yesterday | MPL-2.0 | C | 214 | Claude Code |
| imagegenopenai/codex | Skill | 136k | today | Apache-2.0 | A | 113 | Codex |
| local-ai-agentsmicrosoft/ai-agents-for-beginners | Skill | 80k | yesterday | MIT | A | 200 | Claude Code, Codex |
| playwright-climicrosoft/playwright | Skill | 101k | today | Apache-2.0 | A | 19 | Claude Code |
| browser-automationopenclaw/openclaw | Skill | 395k | today | MIT | A | 36 | Claude Code |
| free-llm-apismnfst/awesome-free-llm-apis | Skill | 14k | today | CC0-1.0 | A | 113 | Claude Code, Codex |
| next-cache-components-adoptionvercel/next.js | Skill | 144k | today | MIT | A | 95 | Claude Code |