agent-owl CLAUDE.md

agent-owl CLAUDE.md is an instructions file for coding agents from arnwaldn/agent-owl. It costs 1,120 tokens per session, scanned A, original, MIT.

A CLAUDE.md file is a set of project-specific instructions for Claude Code, the coding agent. This one documents Agent Owl's file-integrity monitoring, audit trail, and EU AI Act compliance system.

In plain words
What is it for?
Use it when developing or maintaining Agent Owl, including its hashing, audit, compliance, validation, and report-generation code.
Why use it?
It gives the agent the architecture and design decisions needed to work safely in the Agent Owl codebase.

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/arnwaldn/agent-owl/claude-md
Clone the repo
git clone --depth 1 https://github.com/arnwaldn/agent-owl

Wrote 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.

agentmods badge for agent-owl CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/arnwaldn/agent-owl/claude-md.svg)](https://agentmods.dev/instructions/arnwaldn/agent-owl/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/arnwaldn/agent-owl/claude-md"><img src="https://agentmods.dev/badge/instructions/arnwaldn/agent-owl/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 1,120 This file is loaded in full into every session.
When invoked 1,120 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 $0.01120 $0.01120
Opus 5 $0.00560 $0.00560
Sonnet 5 $0.00224 $0.00224
Haiku 4.5 $0.00112 $0.00112

Measured 3d ago against content hash 3805b490be8c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

agent-owl 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.

CLAUDE.md · 106 lines

How it starts

The opening of the file, as written. The whole thing — 106 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agent Owl v3.0

Cryptographic file integrity agent with OWL/RDF audit trail and EU AI Act (Reg. 2024/1689) compliance. Multi-project, fully autonomous — zero configuration needed.

Architecture

agent_owl/
├── ontology.ttl         # OWL 2 DL ontology (TBox) — 709 triples
├── ontology-shacl.ttl   # SHACL shapes — 194 triples (core + high_risk)
├── store.py             # Thread-safe RDF store (rdflib, ReentrantLock)
├── hasher.py            # Streaming hash: SHA-256, SHA-512, BLAKE2b
├── agent.py             # Core agent: watchdog, scan, events, config resolution
├── discovery.py         # Walk-up config finder, project root detection, AgentCache
├── compliance.py        # EU AI Act engine: register, report, retention, incidents
├── validator.py         # SHACL validation (core + high_risk tiers)
├── annexe_iv.py         # Annex IV documentation completeness (9 points)
├── report.py            # HTML/Markdown report generator (Jinja2)
├── _utils.py            # Internal utilities
├── __init__.py          # Package init (v3.1.0)
├── __main__.py          # CLI entry point
└── templates/
    ├── report.html.j2   # HTML compliance report template
    └── report.md.j2     # Markdown compliance report template

agent_owl_mcp_server.py       # FastMCP server — 15 tools, multi-project via AgentCache
hooks/                   # Claude Code hooks (SessionStart, PostWrite, PostBash)
commands/owl-audit.md   # Skill command /owl-audit
audit_store/             # Runtime data (per-project, gitignored)
└── audit.ttl            # Accumulated RDF data (ABox)

Key Design Decisions

  • Config-relative paths: store_path and watch_paths resolve from config file directory, not CWD
  • Immutable TBox: Ontology (ontology.ttl) is loaded once and never modified at runtime
  • Mutable ABox: Audit data (audit.ttl) accumulates events, versions, hashes
  • Thread safety: ReentrantLock in AuditStore, Lock + OrderedDict in AgentCache
  • Snapshot-then-flush: Cache operations snapshot under lock, flush I/O outside lock
  • Auto-init: Hooks detect project roots (15 markers) and initialize AO automatically

Read the full file on GitHub · 106 lines

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. 3d ago First seen · 106 lines · 1,120 tokens per session scan A 3805b490be8c

Subscribe to this mod's changes

agent-owl CLAUDE.md is an instructions file published in the GitHub repository arnwaldn/agent-owl (0 stars, last pushed 1mo ago), licensed MIT. It adds 1,120 tokens to every session, about $0.0056 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.