Borrowing it
Nothing to install: this file belongs to jonapoul/aktual. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/jonapoul/aktual/main/CLAUDE.mdgit clone --depth 1 https://github.com/jonapoul/aktualWrote 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/jonapoul/aktual/claude-md)<a href="https://agentmods.dev/instructions/jonapoul/aktual/claude-md"><img src="https://agentmods.dev/badge/instructions/jonapoul/aktual/claude-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/jonapoul/aktual/claude-md"><img src="https://agentmods.dev/badge/instructions/jonapoul/aktual/claude-md.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.03788 | $0.03788 |
| Opus 5.5 | $0.01515 | $0.01515 |
| Sonnet 5.5 | $0.00758 | $0.00758 |
| Haiku 4.5 | $0.00379 | $0.00379 |
Grade A, and why
aktual 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 — 158 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
Guidance for Claude Code when working in this repo.
Project
Aktual is an unofficial Kotlin Multiplatform client for Actual personal budgeting software, targeting Android and Desktop (JVM). Pet project, unaffiliated with upstream.
General
- IMPORTANT: After any architectural/structural change, grep
**/CLAUDE.mdfor references that need updating (scopes, annotations, module paths, DI patterns). Don't wait to be asked. - Limit visibility where possible: make a type
internalunless another module uses it. Give injected classes (@Inject,@AssistedInject,@Contributes*) aninternal constructorunless another module calls it, so their dependencies stay out of the public API. Metro still wires them across modules. - Prefer
kotlinx.immutablecollections in the UI layer, not plainList/Set. - Prefer
stateFlow.update { x }overstateFlow.value = x. - In coroutine code, catch failures with
runCatchingCancellable(attempt = { }, onFailure = { e -> }, finally = { })(aktual-core), notrunCatchingor a hand-writtencatch (e: CancellationException) { throw e }. It rethrows cancellation and returns the result ofattempt, or ofonFailureif it threw. Usefinallyinstead of wrapping it intry/finally. - The project compiles with
-Xcontext-sensitive-resolution(CSR). Write enum entries, sealed subtypes and companion properties unqualified wherever the expected type is known, and don't import the entry. TheRedundantQualifierdetekt rule flags any you miss. Don't qualify just in case.// Unqualified takesRole(Admin); role == Admin; val role: Role = Admin; assertThat(role).isEqualTo(Admin) when (role) { Admin -> 1 }; state is Failure; state as Failure; takesKind(Button); listOf<Role>(Admin, Basic) role ?: Admin; amount > Zero; Admin in roles; val f: () -> Role = { Admin } // Must qualify: no expected type, or a constructor call val role = Role.Admin; listOf(Role.Admin); a to Role.Admin; roles += Role.Admin val r = if (x) Role.Admin else Role.Basic; takesState(State.Failure("x"))- Only a type, function or imported property in scope makes a name ambiguous (an icon import like
aktual.core.icons.Copyclashes withLogExportTarget.Copy) (LoginMethod.Passwordvs thePasswordclass,Role.Checkboxvs theCheckboxcomposable). Another enum's entry with the same name doesn't count, soconsume(Plus)andUnaryNode(Plus, x)both resolve. - CSR can't infer a type argument from the name, so it doesn't apply where the parameter type mentions the call's own type parameter, like
get(key: DbMetadataKey<T>). Statically importDbMetadataKeycompanion keys instead (import aktual.budget.model.DbMetadataKey.Companion.BudgetName, thenprefs[BudgetName]). - Don't name new enum entries or sealed subtypes after Kotlin types (
Number,Any,Boolean,Error), since that forces a qualifier at every use.
- Only a type, function or imported property in scope makes a name ambiguous (an icon import like
Strings.xyz(user-facing text) is generated from XML inaktual-core:l10n- add the string there and regenerate, don't hardcode. See aktual-core/l10n.- Before using or adding an icon, check upstream Actual for an equivalent and prefer it over Material. See aktual-core/icons.
- In tests, observe
Flow/StateFlowemissions with Turbine (flow.test { awaitItem() }), not by reading.valueor manual collectors. - Put tests in
commonTest. In modules with an android target they run on both targets. UseandroidHostTestonly for tests needing real Android behaviour (Context,Bundle,Looper, Compose UI), with@RunWith(RobolectricTestRunner::class). Host tests run on plain JUnit otherwise, with Android stubs returning defaults (isReturnDefaultValues). - Wrap comments at 100 columns, counting the indent and the
//prefix (an 8-space indent leaves 89 characters of text). That is wherektfmt --google-stylewraps them, not themax_line_lengthof 120 in.editorconfig. ktfmt doesn't reflow comments: it moves the overflow of each long line onto a new//line of its own, leaving stubs like// Existingmid-paragraph. When writing or editing a multi-line comment, rewrap the whole paragraph so every line fits. A line even one column over gets split, so measure rather than eyeball it: before running ktfmt,git diff -U0 origin/main -- '*.kt' '*.kts' | grep -E '^\+\s*//' | awk 'length > 101'must print nothing (101 allows for the diff's+). After ktfmt the stubs are short enough to pass that check, so look for them instead: a//line of one or two words followed by another//line.
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 Changed · +99 tokens per session 9e53c9a62c95
- 5d ago Changed · +6 lines · +472 tokens per session 6fdc585a3e7d
- 6d ago Changed · +26 tokens per session 1ba74cb2217c
- 8d ago Changed · +1 lines · +38 tokens per session 3079300bf530
- 9d ago Changed · +1 lines · +74 tokens per session e2a4e231794b
- 11d ago Changed · +9 lines · +391 tokens per session 430a81bea77f
- 15d ago Changed · +2 lines · +81 tokens per session df65156ab658
- 17d ago Changed · +4 tokens per session 8525584bf0e6
- 19d ago Changed · +8 tokens per session c7ef41ff5fa4
- 20d ago First seen · 139 lines · 2,595 tokens per session scan A e155cefd2c00
aktual CLAUDE.md is an instructions file published in the GitHub repository jonapoul/aktual (43 stars, last pushed yesterday), licensed Apache-2.0. It adds 3,788 tokens to every session, about $0.0152 per session on Opus 5.5. 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 2026-09-01.
aktual CLAUDE.md compared with instructions from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| aktual CLAUDE.mdjonapoul/aktual | Instructions file | 43 | yesterday | Apache-2.0 | A | 3,788 | Claude Code |
| vscode buildNext.instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 9,719 | GitHub Copilot |
| next.js AGENTS.mdvercel/next.js | Instructions file | 143k | today | MIT | A | 7,461 | Codex, OpenCode |
| langchain AGENTS.mdlangchain-ai/langchain | Instructions file | 149k | yesterday | MIT | A | 4,680 | Codex, OpenCode |
| vscode oss-third-party-notices.instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 5,516 | GitHub Copilot |
| vscode copilot-instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 3,093 | GitHub Copilot |
| deepseek-harness AGENTS.mddeepseek-ai/deepseek-harness | Instructions file | 274k | 2d ago | MIT | A | 4,313 | Codex, OpenCode |