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/theagenticguy/erpaval/claude-mdgit clone --depth 1 https://github.com/theagenticguy/erpavalWrote 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/theagenticguy/erpaval/claude-md)<a href="https://agentmods.dev/instructions/theagenticguy/erpaval/claude-md"><img src="https://agentmods.dev/badge/instructions/theagenticguy/erpaval/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 | $0.00562 | $0.00562 |
| Opus 5 | $0.00281 | $0.00281 |
| Sonnet 5 | $0.00112 | $0.00112 |
| Haiku 4.5 | $0.00056 | $0.00056 |
Grade A, and why
erpaval 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 4d 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 — 50 lines — stays where its author put it; the contents beside it link to each section on GitHub.
erpaval-plugin
Development context for the erpaval-plugin repo. This CLAUDE.md is project-level memory — it is NOT shipped as a plugin component.
Repo structure
.claude-plugin/plugin.json # plugin manifest
.mcp.json # MCP servers used by the researcher agent
skills/
erpaval/ # the core skill
SKILL.md
references/ # flow, classifiers, glossary, orchestrator, etc.
templates/ # session YAML/MD skeletons, lesson templates, EARS/HMW stubs
tools/ # PEP 723 scripts: erpaval-new, erpaval-recall, erpaval-validate
product-discovery/ # hard dependency — HMW and EARS substeps
references/frameworks/ # ears.md, how-might-we.md
references/roles/ # hmw-framer.md, ears-specifier.md
templates/ # hmw-skeleton.md, ears-spec-skeleton.md
agents/
researcher.md # Research-phase agent (context7, deepwiki, brave, exa)
hooks/
framework.py # Pydantic hook framework — do not edit without testing
hooks.json # hook event bindings
session_start_bootstrap.py # SessionStart — emits prior-lesson summary
validate_packet.py # PostToolUse(Write|Edit) — Pydantic-checks .erpaval/ writes
compound_nudge.py # Stop — one-shot nudge when Compound is pending
Version management
Bump plugin.json .version manually or with mise run bump -- <major|minor|patch> if you add a mise.toml.
Tools
All three tools under skills/erpaval/tools/ are PEP 723 scripts. Run with uv run. No venv needed.
erpaval-new.py --request "<text>"— scaffold a new session direrpaval-recall.py bootstrap— session-start summaryerpaval-recall.py search --module <path> --tags <csv>— per-task retrievalerpaval-validate.py <path>— schema-check a YAML packet
Hooks
All hooks are fail-open. A broken hook exits 0 and logs to stderr — it cannot wedge a session. The framework.py module handles stdin/stdout plumbing and HookState for session-scoped persistence.
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.
- 4d ago First seen · 50 lines · 562 tokens per session scan A c715ce74638e
erpaval CLAUDE.md is an instructions file published in the GitHub repository theagenticguy/erpaval (27 stars, last pushed 3mo ago), licensed MIT. It adds 562 tokens to every session, about $0.0028 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 2026-08-30.
Other instructions, from other repositories
skill-engine CLAUDE.md
Instructions for nick-railsback/skill-engine, covering skill-engine — standing rules, before proposing a commit, versions and git.
hplan CLAUDE.md
Instructions for kimsanguine/hplan, covering claude.md — hplan project rules (adk l1: memory), hplan context, 9 behavioral rules, rule 1 — think before coding and rule 2 — simplicity first.
ed3d-plugins CLAUDE.md
Claude Code instructions for ed3dai/ed3d-plugins, covering ed3d-plugins, conventions, task invocations use xml syntax and version updates require marketplace and changelog sync.
yoink AGENTS.md
Instructions for theogbrand/yoink, a project described as: yoink is an AI agent that removes complex dependencies by reimplementing only what you need.
tripod CLAUDE.md
Claude Code instructions for joryeugene/tripod, covering priority 0: the tripod, 1. antifragile, 2. simple (hickey: complecting), 3. research first (stand on giants) and priority 1: the contract.
beadle AGENTS.md
Instructions for punt-labs/beadle, covering agent instructions, quick reference and landing the plane (session completion).