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/itlackey/akm/agents-mdgit clone --depth 1 https://github.com/itlackey/akmWhat 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 | $0.02755 | $0.02755 |
| Opus 5 | $0.01378 | $0.01378 |
| Sonnet 5 | $0.00551 | $0.00551 |
| Haiku 4.5 | $0.00276 | $0.00276 |
Grade A, and why
akm 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 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 — 60 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AKM Repo Notes
Workflow
- Runtime and tooling are Bun-first. Use
bun install. - CI runs
bun run check, which isbun run lint && bunx tsc --noEmit && bun run test:unit && bun run test:integration. - Before committing, run
bunx biome check --write src/ tests/. Repo guidance prefers the write-capable Biome pass, not justbun run lint. - Build with
bun run build. It compilessrc/**only intodist/;dist/testsshould never appear. - Prefer focused verification with
bun test tests/<file>.test.ts.bun run check:changedruns a small set of output/contract suites (output-baseline, registry-search, show-argv-entrypoint, output-shapes-unit) plusbun run lintandbunx tsc --noEmit.
Architecture
- This is a CLI-only package. There is no public API, no barrel exports, and no
exportsmap.src/cli.tsis the thin dispatcher; command implementations live undersrc/commands/, mostly in per-family directories (src/commands/read/,src/commands/improve/,src/commands/sources/,src/commands/env/,src/commands/tasks/,src/commands/agent/,src/commands/proposal/,src/commands/health/,src/commands/lint/), with a handful of standalone*-cli.tsfiles (e.g.config-cli.ts,workflow-cli.ts,migrate-cli.ts,registry-cli.ts) still at the top level. - If you touch providers, refs, search/show behavior, config, or output shaping, read
docs/architecture/architecture.mdfirst.tests/contracts/pins active contracts and is meant to catch contract drift. - Supported source providers are locked to
filesystem,git,website, andnpm. Do not addcontext-hub; do not reintroduceopenviking. SourceProvideris exactly{ name, kind, path, sync? }. All providers materialize files to local disk.- Asset refs are
[bundle//]conceptId[#fragment], whereconceptIdis subdir-qualified within its bundle (e.g.skills/code-review,memories/vpn-note,knowledge/api-guide,env/prod). Durable state stores the fully-qualifiedbundle//conceptId; the short bundle-omitted form is input sugar resolved againstdefaultBundle, then the remaining bundles in installation-priority order. Source locators likegithub:owner/repoare forakm bundle add, not for asset addressing. The old[origin//]type:namegrammar is gone (the frozen migrator inscripts/akm-migrate/migrate/is the only place it survives). showis local-index only: resolve through the FTS index, then read from disk. No per-providershowexists.- Registry results are opt-in, stay separate from normal stash hits, and live in
registryHits, neverhits. - All write-target branching by
source.kindbelongs insrc/core/write-source.ts. - Write-target resolution order is
--target->defaultWriteTarget-> working stash (defaultBundle); there is no fallback to the first writable source. writabledefaults totrueonfilesystemandfalseongit/website/npm;writable: trueonwebsiteornpmis rejected at config load.
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 · 60 lines · 2,755 tokens per session scan A c96ba2da30b9
akm AGENTS.md is an instructions file published in the GitHub repository itlackey/akm (55 stars, last pushed 2d ago), licensed MPL-2.0. It adds 2,755 tokens to every session, about $0.0138 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
skills AGENTS.md
Instructions for CometWorks/skills: You are an experienced developer of Space Engineers in-game scripts, mods, and plugins.
skills copilot-instructions.md
Instructions for CometWorks/skills: Carefully read and internalize AGENTS.md from the repository's root before doing anything else.
adeptability AGENTS.md
Instructions for itaywol/adeptability, a project described as: Cross-harness AI skill portability CLI. Author an agent skill once, sync it into Claude Code, Cursor, Copilot, Codex, OpenCode & 45+ AI coding agents. Safety scanner, content-hash drift detection. A package manager / dotfiles for AI coding agent skills.
ZackEyes AGENTS.md
Instructions for yangshiqi/ZackEyes, covering agents.md, 开发流程, 反馈循环, 编译验证 and 编译全部.
ZackEyes CLAUDE.md
Instructions for yangshiqi/ZackEyes, covering claude.md, 上下文分层, build & dev commands, 编译 and 运行.
ux-ui-agent-skills CLAUDE.md
Claude Code instructions for sterlingpathological200/ux-ui-agent-skills, covering ux/ui expert agent — claude design system skill, decision framework, design principles, atomic design and design thinking.