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/hahwul/gori/agents-mdgit clone --depth 1 https://github.com/hahwul/goriWhat 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 | $0.03943 | $0.03943 |
| Opus 5 | $0.01972 | $0.01972 |
| Sonnet 5 | $0.00789 | $0.00789 |
| Haiku 4.5 | $0.00394 | $0.00394 |
Grade A, and why
gori AGENTS.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 — 242 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md: working on gori
gori is an intercepting proxy and workbench for authorized security testing, written in Crystal and shipped as one binary. It sits in the loop between a client and its target, capturing every request and response as a flow you can intercept, replay, fuzz, and scan across HTTP/1.1, HTTP/2, WebSocket, gRPC, and SSE.
Three entry points, one engine layer underneath: gori (TUI), gori mcp (stdio JSON-RPC for
agents), gori run <sub> (headless, for scripts).
This file is the short version, and it is the whole contract for a change that stays inside
one subsystem. DESIGN.md is the long one, and its numbering is
load-bearing: source comments cite principles as (P4), (P6/P7) and sections as
DESIGN.md §4.
- Changing behavior? Read Invariants below first — those three are what changes get wrong.
- Committing? House rules has the commit, CHANGELOG and pre-commit checklist.
- Adding a subsystem? DESIGN.md, then back here.
Invariants: three things not to get wrong
1. Malformed input is the payload (P7)
A security proxy that sanitizes its operator's bytes is broken. The captured wire bytes are canonical; parsed columns and pretty views are derived projections.
- There is no stdlib
HTTP::Requestparsing anywhere on the proxy path.src/gori/proxy/codec/http1.cris a byte-exact sans-IO codec: it flagsmalformed?and keeps the octets rather than rejecting.serialize_headis the identity function. - The axis is provenance, not byte values. The same octet gets three different answers:
- operator bytes (imported HAR, an MCP
rawrequest, a replay) go out verbatim, never sanitized. See the comment atsrc/gori/import/builder.cr:31-38andsrc/gori/mcp/request_builder.cr(normalize_raw). - page-authored bytes (a crawled
<a href>) get percent-encoded where they merely break, and refused where they frame. Seesrc/gori/discover/url.cr(encode_unsafe). - remote-chosen bytes (a redirect
Location) are refused outright, not repaired. Seesrc/gori/fuzz/engine.cr(redirect following).
- operator bytes (imported HAR, an MCP
- The predicate has one home:
Codec::Http1.request_token_safe?. Do not re-derive it next to a new caller; that exact shape has already recurred three times (#390, #394, #397). - Request framing rejects any obfuscation (
obfuscated_header?); response framing is deliberately narrower (framing_ambiguous?, and read the comment above it). Do not "symmetrize" them: a request's peer is the operator's own browser, a response's peer is the whole internet.
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 · 242 lines · 3,943 tokens per session scan A e815f3495fcf
gori AGENTS.md is an instructions file published in the GitHub repository hahwul/gori (85 stars, last pushed 3d ago), licensed Apache-2.0. It adds 3,943 tokens to every session, about $0.0197 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
metabigor CLAUDE.md
Instructions for j3ssie/metabigor, covering metabigor - claude.md, project overview, architecture, directory structure and key design patterns.
metabigor AGENTS.md
Instructions for j3ssie/metabigor, a project described as: OSINT power without API key hassle.
mockserver-monorepo AGENTS.md
AGENTS.md instructions for mock-server/mockserver-monorepo, covering mockserver — agent instructions, instruction priority, project overview, local development environment and project documentation.
plano CLAUDE.md
Instructions for katanemo/plano, covering claude.md, build & test commands, rust — wasm plugins (must target wasm32-wasip1), rust — brightstaff binary (native target) and rust — tests, format, lint.
GreenTunnel CLAUDE.md
Instructions for SadeghHayeri/GreenTunnel, covering greentunnel v3, commands, running the cli, ⚠️ npm run dev changes your real system proxy and layout.
claude-tap AGENTS.md
Instructions for liaohch3/claude-tap, covering maintainer automation notes, agents 索引, documentation boundary, review guidelines and pre-commit hook.