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.
npx agentmods add instructions/ubcent/velar/claude-mdgit clone --depth 1 https://github.com/ubcent/velarWrote 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/ubcent/velar/claude-md)<a href="https://agentmods.dev/instructions/ubcent/velar/claude-md"><img src="https://agentmods.dev/badge/instructions/ubcent/velar/claude-md.svg" alt="Measured on agentmods" 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 | $0.01154 | $0.01154 |
| Opus 5 | $0.00577 | $0.00577 |
| Sonnet 5 | $0.00231 | $0.00231 |
| Haiku 4.5 | $0.00115 | $0.00115 |
Grade A, and why
velar 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 3d 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.
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.
How it starts
The opening of the file, as written. The whole thing — 102 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.
Project Overview
Velar (formerly PromptShield) is a local HTTP/HTTPS proxy with MITM support for AI traffic. It detects and masks PII/secrets in outbound requests to AI providers, then restores original values in responses. All processing is local — nothing is sent to external services.
Commands
# Build both binaries (velar CLI + velard daemon)
make build
# Run all tests with race detector
make test
# Run a single package's tests
go test ./internal/sanitizer/... -run TestSanitize
# Run a single test
go test ./internal/sanitizer/... -run TestSanitize/email -v
# Test ONNX NER detector (requires Python venv with numpy + onnxruntime)
make test-ner
# Run the proxy directly (without daemonizing)
go run ./cmd/velar daemon
# Start as background daemon (requires make build first)
./bin/velar start
Architecture
The project produces two binaries:
cmd/velar— CLI that manages the daemon lifecycle (start,stop,restart,status,logs,stats,ca,proxy,model)cmd/velard— The actual daemon process spawned byvelar start
Request Flow
Client -> Proxy.handle() -> policy.Engine.Evaluate()
-> if CONNECT + MITM domain: mitm.Handler.HandleMITM()
-> if CONNECT + tunnel: handleTunnel()
-> if HTTP: handleHTTP()
-> Inspector.InspectRequest() [sanitize outbound]
-> upstream RoundTrip
-> Inspector.InspectResponse() [restore inbound]
Key Packages
| Package | Role |
|---|---|
internal/proxy |
Core HTTP proxy; routes to MITM or tunnel based on policy |
internal/proxy/mitm |
MITM handler: TLS intercept via per-host leaf certs signed by local CA; uses singleConnListener pattern |
internal/sanitizer |
SanitizingInspector implements the Inspector interface; JSON-aware masking with fallback to full-text |
internal/detect |
Detection backends: RegexDetector (fast), ONNXNERDetector (Python subprocess), HybridDetector (combines both) |
internal/policy |
Rule engine: evaluates allow/block/mitm decisions per host |
internal/session |
Per-request store mapping placeholders -> originals, keyed by session ID propagated via context |
internal/config |
YAML/JSON config loader with custom lite parser (no external YAML dep); env overrides VELAR_PORT, VELAR_LOG_FILE |
internal/audit |
JSONL audit logger for request/response entries |
internal/stats |
In-memory stats collector for velar stats command |
internal/classifier |
Host classifier (stub — used as hook for future classification) |
internal/notifier |
macOS NSUserNotification wrapper (no-op on other platforms) |
internal/systemproxy |
macOS networksetup wrapper to toggle system HTTP/HTTPS proxy |
internal/models |
ONNX model registry and downloader for velar model commands |
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.
- 3d ago First seen · 102 lines · 1,154 tokens per session scan A 5e35163f9d0d
velar CLAUDE.md is an instructions file published in the GitHub repository ubcent/velar (10 stars, last pushed 1mo ago), licensed MIT. It adds 1,154 tokens to every session, about $0.0058 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.
Other instructions, from other repositories
pentestcode CLAUDE.md
Instructions for s0ld13rr/pentestcode, covering claude.md, what this is, project structure, multi-agent architecture and key files.
claude-bug-bounty CLAUDE.md
Instructions for Mikacr1138/claude-bug-bounty, covering claude bug bounty — plugin guide, what's here, commands (8 slash commands), agents (5 specialized agents) and rules (always active).
pentestcode AGENTS.md
Instructions for s0ld13rr/pentestcode, covering branch names, commits and pr titles, style guide, general principles and destructuring.
claude-skills CLAUDE.md
Instructions for mastepanoski/claude-skills, covering claude.md, repository purpose, testing skills locally, repository architecture and core structure.
guardvibe CLAUDE.md
Instructions for goklab/guardvibe, covering project guidelines, guardvibe security scanner, when user asks for security scan/fix, never do these and quick reference.
aileron CLAUDE.md
Claude Code instructions for aileron-sh/aileron, covering claude.md - aileron, the four invariants, the proxy is the enforcement path - treat it accordingly, layout and working here.