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/arslanr-com/amcli/claude-mdgit clone --depth 1 https://github.com/arslanr-com/amcliWhat 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 | $0.05215 | $0.05215 |
| Opus 5 | $0.02608 | $0.02608 |
| Sonnet 5 | $0.01043 | $0.01043 |
| Haiku 4.5 | $0.00522 | $0.00522 |
Grade A, and why
amcli 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.
How it starts
The opening of the file, as written. The whole thing — 327 lines — stays where its author put it; the contents beside it link to each section on GitHub.
amcli — notes for agents working on this repo
amcli is a CLI over ArchiMate model files. Rust workspace, one static binary,
no runtime dependencies.
The rule that overrides everything else
A parse-then-write round trip must be byte-identical. If a change makes
tests/corpus fail the identity test, the change is wrong — not the test. The
whole product proposition is that editing a model produces a diff a human can
review, and that only holds if untouched bytes stay untouched.
Practically: never re-serialize what you did not edit. amcli-xml blits
untouched subtrees straight out of the source buffer, and any new code that
touches the document must preserve that property.
Layout
| Crate | Job |
|---|---|
crates/amcli-xml |
Format-preserving XML tree with byte spans. Knows nothing about ArchiMate. |
crates/amcli-model |
The ArchiMate IR: types, folders, concepts, views, containers (plain XML or the ZIP an embedded image makes of it). |
crates/amcli-view |
View geometry, layout, notation — and icons.rs, the type icons. |
crates/amcli-render |
A compiled view to SVG, or to PNG through resvg (pure Rust, so the binary stays static). |
crates/amcli-cli |
The binary; src/web/ is amcli web, the read-only viewer. |
xtask |
Codegen from the vendored Archi assets. |
skills/amcli/ is shipped, not documentation
npx skills add arslanr-com/amcli copies that directory out of the default
branch verbatim, and crates/amcli-cli/src/skill.rs embeds the same files with
include_str! for the binary-first route. Both routes must produce identical
bytes — both_install_routes_ship_the_same_bytes walks the directory and
fails if a file is there but not in FILES.
Consequences worth knowing before you edit it:
- Nothing may be generated into it. A file written only by
skill installmakes the two routes differ; a committed generated file goes stale on the next release. That is why there is noreferences/commands.mdand whyamcli skill commandsexists instead. - The executable bit does not survive. The skills CLI has two install
paths, and its blob fast path writes every file 0644. Invoke scripts as
sh scripts/install.sh, never./scripts/install.sh. - The skill ships from the branch, the binary from the newest tag, so the
skill is normally the newer of the two. Never tell an agent to reconcile
that by reinstalling the skill — it would downgrade itself. The reconciliation
lives in
parse_or_hintinmain.rs. - Every release updates the skill, and the version line is only half of it.
"written for amcli X.Y.Z" in
SKILL.mdis pinned to the workspace version by a test, so a release commit bumpsCargo.toml,Cargo.lockand that line orcargo testis red. The half no test can check is the prose: a command or flag the release adds is invisible to an agent untilSKILL.mddescribes it, because that file — not--help— is what it reads. So before tagging, diff the release againstamcli skill commandsand ask of each new subcommand, flag and batch op: is it inSKILL.md, or inreferences/batch.mdif it only matters inside a batch? Shipping a feature nobody is told about is the same as not shipping it. - The skill's Setup runs the installer every session. That is only fine
because
install.sh/install.ps1short-circuit when the newest release is already installed and keep the installed binary when there is no network. Keep those two properties if you touch the installers. - Files beginning with a dot are never copied, so no
.gitattributesor.versioninside the skill folder.
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.
- 2d ago First seen · 327 lines · 5,215 tokens per session scan A e66d42d1e6b0
amcli CLAUDE.md is an instructions file published in the GitHub repository arslanr-com/amcli (2 stars, last pushed 3d ago), licensed Apache-2.0. It adds 5,215 tokens to every session, about $0.0261 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
mcp-archimate CLAUDE.md
Claude Code instructions for byrondelgado/mcp-archimate, covering claude.md, project overview, commands, architecture and registration by import side effect.
mcp-archimate AGENTS.md
AGENTS.md instructions for byrondelgado/mcp-archimate, covering agents.md, instruction precedence, canonical instructions, backlog rule and codex translation rules.
togaf10_EAP_MCP CLAUDE.md
Claude Code instructions for Waganawa-Megumin/togaf10_EAP_MCP, covering claude.md, 現在の状態と進め方, この製品が解こうとしている問題, 実装上の約束事(レビューで実際に問題になった点) and コマンド.
chrono AGENTS.md
AGENTS.md instructions for projectchrono/chrono, covering repository guidelines, developers vs. users, project structure & module organization, build, test, and development commands and physics manuals & demo-first workflow.
PINA AGENTS.md
Instructions for PINA-org/PINA, covering pina — physics-informed neural architectures, quick reference, workflow: problem → model → solver → trainer, problem types and condition types.
Thermodynamics.jl AGENTS.md
Instructions for CliMA/Thermodynamics.jl, covering thermodynamics.jl agent guide, ecosystem guidelines, before you act: agent autonomy, repo-specific guidelines and architecture.