Borrowing it
Nothing to install: this file belongs to akanthed/SecureAI-Scan. 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/akanthed/SecureAI-Scan/main/CLAUDE.mdgit clone --depth 1 https://github.com/akanthed/SecureAI-ScanWrote 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/instructions/akanthed/secureai-scan/claude-md)<a href="https://agentmods.dev/instructions/akanthed/secureai-scan/claude-md"><img src="https://agentmods.dev/badge/instructions/akanthed/secureai-scan/claude-md/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/instructions/akanthed/secureai-scan/claude-md"><img src="https://agentmods.dev/badge/instructions/akanthed/secureai-scan/claude-md.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 | $0.05083 | $0.05083 |
| Opus 5.5 | $0.02033 | $0.02033 |
| Sonnet 5.5 | $0.01017 | $0.01017 |
| Haiku 4.5 | $0.00508 | $0.00508 |
Grade C, and why
SecureAI-Scan CLAUDE.md 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 23d ago.
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.
- **`src/scanner/deobfuscate.ts`** — content checks match against normalized *variants* (invisible-stripped, homoglyph-folded, string-splice-joined, intra-word-break-joined) instead of one byte sequence. `matchAcrossVari Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.
Unrestricted tool accessmediumExcessive agency
A wildcard tool grant or "run any command" leaves no least-privilege boundary at all.
- **Report only an intent the code fails to keep.** MCP013 needs the argument *composed into a fixed command*; MCP014 needs it *joined onto a base directory*. A by-design "run any command" or "read any file" tool is `heu How it starts
The opening of the file, as written. The whole thing — 151 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
What this is
secureai-scan (npm) — a static AI/LLM security scanner for TypeScript, JavaScript, and Python. It finds prompt injection, MCP supply-chain risks, RAG poisoning, Agent Skill poisoning, and related OWASP LLM/ASI/MCP Top 10 issues, and proves findings with source→flow→sink dataflow traces rather than keyword matching alone.
Hard requirements (non-negotiable, apply to every change)
1. Zero tolerance for false positives
This is the product's entire value proposition (see the "precision contract" in README.md) — a scanner nobody trusts because it cries wolf is worse than no scanner. Any change to detection logic (a rule file, tool-poisoning-checks.ts, python-scanner.ts, confidence.ts, etc.) must not introduce new false positives, and existing false positives found during work should be fixed, not worked around.
- The internal precision gate (
test-fixtures/safe/+test/corpus.test.js) enforces zeroproven/likelyfindings on known-safe patterns — this must stay green, and any bug fixed for a false positive gets a permanent fixture there so it can't regress silently. - That corpus alone is not sufficient — it only proves the scanner behaves on code we wrote to test it. See the regression-scan requirement below for the check that catches what it can't.
- When you find a false positive, fix the root cause (a rule's evidence tier, an unguarded keyword match, a missing
isTestFilePathdemotion, an overly broad name set) rather than special-casing the one input that revealed it.
2. Test against real-world repos, not just fixtures, before calling detection work done
For any change that touches detection logic (new rule, edited rule, edited shared matcher), run:
npm run regression
This clones a curated set of real public repos (scripts/regression-scan.js — OpenAI/Anthropic/Vercel AI SDKs, official MCP servers and SDK, LlamaIndex) into .regression-cache/ (gitignored, cached across runs — pass --fresh to re-clone, or a repo name to scan just one) and scans each with the built CLI. It is a gate: it exits non-zero on any proven/likely finding whose repo|rule|file fingerprint isn't in test/regression-baseline.json. Fingerprints omit line numbers so ordinary upstream churn isn't noise. Read every new finding against its source line:
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.
- 23d ago Changed · +11 lines · +531 tokens per session scan B → C 53aa140f259a
- 1mo ago First seen · 140 lines · 4,552 tokens per session scan B df159c6fce8f
SecureAI-Scan CLAUDE.md is an instructions file published in the GitHub repository akanthed/SecureAI-Scan (24 stars, last pushed yesterday), licensed MIT. It adds 5,083 tokens to every session, about $0.0203 per session on Opus 5.5. A static security scan graded it C with 2 findings (instruction-override phrasing, unrestricted tool access). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, cache eligibility and rollout and applying the notice (cutover).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode copilot-instructions.md
Copilot instructions for microsoft/vscode, covering vs code copilot instructions, project overview, root folders, core architecture (src/ folder) and built-in extensions (extensions/ folder).
vscode model-prompts.instructions.md
Model-specific prompt authoring and registry guidelines.