agentstack CLAUDE.md

Project instructions for AgentStack, a Rust workspace that governs how coding agents use trust, policy, runtime, recording, import, rendering, and restore code. A workspace is a repository containing related packages that are built together.

In plain words
What is it for?
Use them when modifying AgentStack to find the authoritative strategy, work queue, architecture, enforcement rules, and package boundaries, and to guide Rust implementation choices.
Why use it?
They help contributors extend existing code safely and keep enforcement in the command-line interface instead of duplicating or bypassing established paths.

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/tarekkharsa/agentstack/claude-md
Clone the repo
git clone --depth 1 https://github.com/Tarekkharsa/agentstack
Per session 562 This file is loaded in full into every session.
When invoked 562 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 $0.00562 $0.00562
Opus 5 $0.00281 $0.00281
Sonnet 5 $0.00112 $0.00112
Haiku 4.5 $0.00056 $0.00056

Measured 2d ago against content hash f1b7d77b08de, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

agentstack 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 2d 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 · 52 lines

How it starts

The opening of the file, as written. The whole thing — 52 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md — AgentStack

Shipped Rust workspace. Extend existing seams; do not reimplement working trust, policy, gateway, runtime, recording, import, render, or restore paths. Prefer clear, boring, idiomatic Rust; explain non-obvious ownership/lifetime/ trait/error choices briefly.

crates/
  core/       manifest, lockfile, digests
  trust/      content-bound consent and signatures
  policy/     machine-first policy intersection
  adapters/   native config compilers for supported CLIs
  recorder/   call and run evidence
  runtime/    sandbox orchestration
  egress/     enforced network proxy
  executor/   policy-agnostic governed execution domain
  workflow/   self-contained Boa workflow engine
  mcp/        MCP protocol boundary, RMCP confined behind a JSON backend
  cli/        binary, orchestration, JSON/action APIs

Authorities: STRATEGY.md (strategy), TODO.md (the only work queue), docs/ARCHITECTURE.md + docs/ENFORCEMENT.md (what the code does and enforces). docs/archive/ is history, never direction. The CLI is the primary surface and source of authority; t3code is an optional companion calling stable read APIs and fixed actions, never an enforcement boundary. Never recreate a second UI.

Invariants

The non-negotiable invariants live in STRATEGY.md ("What never changes") and docs/ENFORCEMENT.md; every change must preserve them. Hooks always get the full consent ceremony; trust and policy stay small review boundaries; new dependencies need maintainer approval; Boa stays isolated in workflow.

Hard rules — context and tests

  • Never run the full test suite locally. Loop: cargo check -p <crate> while iterating; cargo test -p <crate> (or --test <name>) only for crates the change can break; full suite belongs to CI.
  • Before handoff: cargo check --workspace --all-targets (the only cheap check that sees all test targets), cargo fmt --check, relevant clippy.

Hard rules — build cache

  • Building in .claude/worktrees/*: export CARGO_TARGET_DIR="$HOME/.cache/agentstack-target" first, so every worktree shares one incremental cache instead of recompiling the workspace per tree.
  • Debug builds by default; --release only when CI parity demands it.
  • Removing a worktree: delete its local target/ with it.

Read the full file on GitHub · 52 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. 2d ago First seen · 52 lines · 562 tokens per session scan A f1b7d77b08de

Subscribe to this mod's changes

agentstack CLAUDE.md is an instructions file published in the GitHub repository Tarekkharsa/agentstack (3 stars, last pushed 19d ago), licensed Apache-2.0. 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 2026-08-31.