code-pentester CLAUDE.md

Repository guidance for a code-security testing project that requires findings to include verified evidence from specific files and lines.

In plain words
What is it for?
Testing code for security issues, verifying whether patterns are present or absent, following project rules, and recording focused, reviewable changes.
Why use it?
It helps keep security analysis accurate by separating confirmed facts from assumptions and by checking claims before reporting them.

Instructions file

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/ch015/code-pentester/claude-md
Clone the repo
git clone --depth 1 https://github.com/ch015/code-pentester
Per session 480 This file is loaded in full into every session.
When invoked 480 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00480 $0.00480
Opus 5 $0.00240 $0.00240
Sonnet 5 $0.00096 $0.00096
Haiku 4.5 $0.00048 $0.00048

Measured yesterday against content hash 20c5686334dc, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

code-pentester CLAUDE.md scanned grade A with 0 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 yesterday.

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.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

CLAUDE.md · 27 lines

What it actually says

CLAUDE.md — ch015-pentester

Reasoning Discipline

ch015 자체가 모든 Finding에 Observed Evidence (file:line) 를 요구한다. 같은 원칙을 어시스턴트의 ch015 관련 답변에도 적용한다.

  1. Verify before declaring absence. "X is not in ch015" requires grep -r X skills/ agents/ knowledge-base/ templates/. Distinguish "verified absent" from "haven't seen yet."

  2. Recommend ONE option. 추천이 사실상 하나면 하나만 제시. C × E처럼 같은 답을 두 번 묶지 말 것. 비교표는 실제 trade-off가 의미 있을 때만.

  3. Verify pushback with same rigor. 사용자가 반박하면 grep/read로 그 주장도 검증한 동의/반박. 반사적 동의는 sycophancy.

  4. Cite file:line. 무거운 결론에 file:line 인용 없으면 가설이지 사실이 아님. 인용 없는 단정 금지.

  5. Broad-scope = 2-pass sampling. "How does ch015 do X" 류 질문은 SKILL + agent + template + knowledge 레이어를 모두 본다. 첫 grep은 표본, 결론 직전에 확인 패스 한 번 더.

Project Operational Rules

  • No direct push to main. 로컬 커밋은 OK, push/PR은 명시 요청 시에만.
  • reports/ is gitignored. 벤치마크 출력물은 로컬 전용 산출물.
  • Sub-agents inherit parent plan mode. 부모(어시스턴트)가 plan mode면 자식 에이전트도 plan mode에 갇힘. 자식이 실행 안 되고 플래닝만 반복하면 부모가 ExitPlanMode 먼저 호출하고 재실행.
  • Engagement dirs are under target project (예: /Users/philip/workdir/benchmarks/juice-shop/.ch015-engagement-*), NOT under ch015-pentester repo.

Tone

Korean default. Terse, no preamble, no trailing summary unless requested.

Changes

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.

  1. yesterday First seen · 27 lines · 480 tokens per session scan A 20c5686334dc

Subscribe to this mod's changes

code-pentester CLAUDE.md is an instructions file published in the GitHub repository ch015/code-pentester (4 stars, last pushed 13d ago), licensed MIT. It adds 480 tokens to every session, about $0.0024 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.