surshar / claude-code

Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands.

		@@ -2,7 +2,7 @@
2	2
3	3	A mod is a Claude Code plugin whose behaviour lives in a hooks module: one
4	4	`register(on, options)` entry that hooks the engine's events as functions
5		-`($, e, next)`. These three ship inside Claude Code; this folder is their
	5	+`($, e, next)`. These four ship inside Claude Code; this folder is their
6	6	source, published as it is built into the binary.
7	7
8	8	\| Mod \| What it does \| Seated \|
		@@ -10,6 +10,7 @@
10	10	\| [`sec-default`](sec-default) \| Keeps an organization's classic hooks, prompt content, managed settings and tool policy out of reach of the plugins a person installs; adds no policy of its own. \| Outermost, on a machine with managed settings or for a Team or Enterprise organization, unless managed `prependPlugins` says otherwise \|
11	11	\| [`diff`](diff) \| `/diff`: the session's uncommitted changes in a pane beside the transcript, file by file with their hunks, refreshed as Claude edits files and runs commands. \| Built in \|
12	12	\| [`telemetry`](telemetry) \| Adds `$.telemetry` (`log`, `mark`) in the `engine.create` fold so a plugin can record an event as a first-party analytics row; sends nothing wherever Claude Code's analytics are off. \| Built in \|
	13	+\| [`agents-md`](agents-md) \| `AGENTS.md` as project instructions, by one option: loaded where the project has no `CLAUDE.md` of its own (`claude-md-or-agents-md`, the default) or beside it (`claude-md-and-agents-md`), placed and framed exactly as the engine places `CLAUDE.md`, nested ones on a `Read`; or the project's and the person's instruction files dropped and the organization's kept (`managed-only`); or `CLAUDE.md` alone, as the engine reads it (`claude-md`). \| Built in \|
13	14
14	15	Each folder is a complete plugin: `.claude-plugin/plugin.json`, a
15	16	`hooks/hooks.json` naming the module, and TypeScript under `hooks/` typed

		@@ -1,0 +1,23 @@
	1	+{
	2	+ "name": "agents-md",
	3	+ "version": "0.1.0",
	4	+ "description": "AGENTS.md as project instructions, by one option. Under claude-md-or-agents-md a project with no instruction files of its own gets its AGENTS.md files instead: a prompt.context hook asks $.fs.ancestors for every AGENTS.md and .claude/AGENTS.md from the filesystem root down to the working directory and hands them to the engine as project instruction files, which it renders, announces and withholds from agents exactly as it does CLAUDE.md, skipping any file the engine already loaded by import or link; a tool.call hook on Read attaches the ones under the project root as the engine attaches a nested CLAUDE.md. Under claude-md-and-agents-md every AGENTS.md is loaded beside CLAUDE.md. Under managed-only the project's and the person's instruction files are dropped and the organization's managed ones kept. Under claude-md the engine's walk stands alone. claude-md-or-agents-md is the default.",
	5	+ "author": {
	6	+ "name": "Anthropic"
	7	+ },
	8	+ "userConfig": {
	9	+ "instructionFiles": {
	10	+ "type": "string",
	11	+ "title": "Project instructions",
	12	+ "description": "\"claude-md\": CLAUDE.md only, loaded by the engine as today. \"claude-md-or-agents-md\" (default): a project with no CLAUDE.md of its own gets its AGENTS.md files instead, loaded exactly where and how CLAUDE.md would be. \"claude-md-and-agents-md\": AGENTS.md files are loaded beside CLAUDE.md (a file CLAUDE.md already imports or links to is not loaded twice). \"managed-only\": the project's and your own instruction files are dropped; the organization's managed CLAUDE.md and memory stay.",
	13	+ "required": false,
	14	+ "default": "claude-md-or-agents-md",
	15	+ "options": [
	16	+ "claude-md",
	17	+ "claude-md-or-agents-md",
	18	+ "claude-md-and-agents-md",
	19	+ "managed-only"
	20	+ ]
	21	+ }
	22	+ }
	23	+}

		@@ -1,0 +1,169 @@
	1	+# agents-md
	2	+
	3	+`AGENTS.md` read the way Claude Code reads `CLAUDE.md`, as a plugin, under
	4	+one option, `instructionFiles`:
	5	+
	6	+- `claude-md`: only `CLAUDE.md` is loaded, by the engine, as today. The plugin
	7	+ adds nothing.
	8	+- `claude-md-or-agents-md` (the default): a project with no instruction files
	9	+ of its own gets its `AGENTS.md` files instead, loaded exactly where and how
	10	+ `CLAUDE.md` would be. "Of its own" is read off what the engine loaded for
	11	+ the context: a `CLAUDE.md`, `.claude/CLAUDE.md` or `CLAUDE.local.md` in any
	12	+ directory from the root down to the working directory leaves the whole
	13	+ project to the engine, and the plugin stays out (the organization's managed
	14	+ file, the person's `~/.claude/CLAUDE.md`, a `.claude/rules` file and an
	15	+ added directory's `CLAUDE.md` do not count, as the nested walk does not see
	16	+ them either). With none, every `AGENTS.md` and `.claude/AGENTS.md` on that
	17	+ path joins the instruction files the engine renders, and a `Read` under a
	18	+ subdirectory attaches that directory's `AGENTS.md` unless a `CLAUDE.md`
	19	+ there claims it.
	20	+- `claude-md-and-agents-md`: every `AGENTS.md` is loaded beside `CLAUDE.md`,
	21	+ up and down the tree; a file `CLAUDE.md` already `@`-imports, or is a link
	22	+ to, is not loaded a second time (compared by path, then by content).
	23	+- `managed-only`: the project's checked-in and private instruction files and
	24	+ the person's own are dropped from the context; the organization's managed
	25	+ `CLAUDE.md` and the engine's memory stay. The engine's nested `CLAUDE.md`
	26	+ attachments on `Read` are not an event yet and still arrive. (The engine's
	27	+ `claudeMdExcludes` setting also exists, for user, project and local files,
	28	+ and applies to the `AGENTS.md` files this plugin reads too.)
	29	+
	30	+How the files reach the model is the engine's doing, not the plugin's:
	31	+`prompt.context` hands a hook the instruction files behind `claudeMd`
	32	+(`{ path, kind, content, parent? }`, kinds `managed`, `user`, `project`,
	33	+`local`, `memory`, in load order) and a hook answers the list changed. The
	34	+engine then renders `claudeMd` from the answered files with its own preamble
	35	+and framing, announces them by name, and keeps only the `managed` ones for an
	36	+agent that omits project instructions (Explore, Plan, a custom agent with
	37	+`omitClaudeMd`). So an `AGENTS.md` this plugin adds as a `project` file is, to
	38	+everything downstream, a project instruction file: same place in the context,
	39	+same framing, same omission rules, same announcement. An organization's
	40	+prepended plugin on `prompt.context` sits above this one and has the last
	41	+word on the files.
	42	+
	43	+`hooks/register.ts` is the module; everything under `hooks/` is its parts,
	44	+importing `claude-code` and one another alone. `tests/` runs under
	45	+`claude plugin test <this folder>`.
	46	+
	47	+## Setting the option
	48	+
	49	+As a built-in its option is the `/config` row "Project instructions", a
	50	+picker over the four values, each described there. By hand it is
	51	+
	52	+```json
	53	+{
	54	+ "pluginConfigs": {
	55	+ "agents-md@builtin": {
	56	+ "options": { "instructionFiles": "claude-md-and-agents-md" }
	57	+ }
	58	+ }
	59	+}
	60	+```
	61	+
	62	+in user settings (`~/.claude/settings.json`), `--settings`, or managed
	63	+settings; a project's `.claude/settings.json` is not read for plugin
	64	+options. Changing it reloads the module, and the next context the engine
	65	+builds (the next turn after the reload, a new conversation, `/clear`, a
	66	+compaction) carries the new mode's files. A hand-typed value outside the
	67	+four is told once in the transcript and reads as the default. `/plugin`
	68	+lists the plugin among the built-ins, where a person can turn it off; with
	69	+it off the engine reads `CLAUDE.md` alone. No hooks setting or CLI mode turns
	70	+it off (`disableAllHooks`, `allowManagedHooksOnly` and `--bare` govern
	71	+settings hooks and installed plugins, not built-ins); where the engine loads
	72	+no instruction files (`--bare` without `--add-dir`, `--safe-mode`,
	73	+`CLAUDE_CODE_DISABLE_CLAUDE_MDS`) its walk finds none and it adds none,
	74	+`CLAUDE.md` and `AGENTS.md` alike.
	75	+
	76	+The option was first keyed `projectInstructions`, with the values `claude`,
	77	+`agents-fallback`, `both` and `none`. A value still stored under that key is
	78	+honoured for now while `instructionFiles` reads as its default: `none` as
	79	+`managed-only`, `claude` as `claude-md`, `agents-fallback` as
	80	+`claude-md-or-agents-md`, `both` as `claude-md-and-agents-md`, any other
	81	+value as `claude-md` (which adds nothing, never as the default, which loads
	82	+`AGENTS.md`); the first `session.start` of a load says in the transcript how
	83	+it was read. Once `instructionFiles` is set to anything but its default, the
	84	+old key is not read and the transcript says to remove it.
	85	+
	86	+Run from this folder instead (`claude --plugin-dir mods/agents-md`), the
	87	+same entry is keyed `"agents-md"`.
	88	+
	89	+## What it hooks
	90	+
	91	+\| event \| what the hook does \|
	92	+\| --- \| --- \|
	93	+\| `session.start` \| in every mode: passes the start straight through and floats the usage row for the configured mode, never awaited; the first start of a load logs how a stored `projectInstructions` value is read. The session's start never waits on this plugin \|
	94	+\| `prompt.context` \| under `claude-md-or-agents-md` and `claude-md-and-agents-md`: walks `$.fs.ancestors` for the `AGENTS.md` files above the working directory and answers them as `project` instruction files, each `@` import its own entry after its file, each placed where a project file of its directory stands (root first, before the first deeper project file, else after the last project file, before memory); files the engine already holds by path or by content are left out; under `claude-md-or-agents-md` it answers nothing when the project has a `CLAUDE.md` of its own (among the handed files, else found by a `$.fs.ancestors` walk, so a `CLAUDE.md` the engine loaded and then withheld still counts), and logs which files it loaded once, and again after a move to another project root; handed unknown files (a hook above rewrote the `claudeMd` text) it adds nothing; the first context of a load sends the load row (counts) and the feature mark; under `managed-only` (matcher: a `project`, `local` or `user` file present): answers the list without those kinds \|
	95	+\| `agent.spawn` on `fork: true` \| under `claude-md-or-agents-md` and `claude-md-and-agents-md`: a fork the Agent tool starts shares its parent's prompt prefix, so the parent loop's delivered nested files are copied to the fork's loop and not attached to it again (a `/fork` or `/subtask` fork does not raise `agent.spawn` yet and starts from an empty set, as every fork did before; a fork started in the same tool batch as a `Read` inherits that Read's file although its prefix holds a placeholder for it) \|
	96	+\| `tool.call` on `Read` \| under `claude-md-or-agents-md` and `claude-md-and-agents-md`, for a file under the session's project root (`$.session.root()`, read live, so `/cd`, a host's directory change and worktree moves are followed and a moved root starts the delivered sets and the fallback decision over; a file elsewhere gets nothing, as the engine attaches no nested `CLAUDE.md` there; and nothing anywhere in a run where the engine attaches nothing to a turn, `--bare` with its `CLAUDE_CODE_SIMPLE` or `CLAUDE_CODE_DISABLE_ATTACHMENTS`, read on every Read through `$.env.get` as the engine reads them on every turn): walks only the directories strictly between the root and the read file (`$.fs.ancestors` with `below: root`, as the engine walks only those for a nested `CLAUDE.md`, never up to the filesystem root again) and attaches their `AGENTS.md` files not yet given to that agent loop, not already among the context's instruction files (by path or, for a project file, by text) and not claimed by a `CLAUDE.md` of the same directory (or imported by one), as `context` after the tool result, framed `Contents of <path>:` byte for byte as the engine frames a nested `CLAUDE.md`, whatever its size; each file once per loop and conversation (the context's recomputation after a compaction or `/clear` starts the count over), the context's files never; a Read that attached files sends the nested row. A `~` or `~/` path is read under the home directory as the Read tool reads it \|
	97	+
	98	+## What it calls on `$`
	99	+
	100	+`fs.ancestors` (with each found file's `parts`: the file and its imports
	101	+apart; with `below` on a Read; it finds nothing on a thin client, whose
	102	+workspace files are remote, as the engine's own walk does), `session.root`,
	103	+`session.cwd`, `env.get` (`HOME` and `USERPROFILE`, once per load, the
	104	+profile first on a Windows spelling of the working directory, so a `~/` path
	105	+the model hands a Read resolves where the Read tool reads it; `CLAUDE_CODE_SIMPLE`
	106	+and `CLAUDE_CODE_DISABLE_ATTACHMENTS` on every Read), `ui.log`,
	107	+`telemetry.log` and `telemetry.mark`.
	108	+
	109	+`$.telemetry` is the [telemetry](../telemetry) plugin's noun; where that
	110	+plugin is not seated the calls find no noun and are dropped without a trace,
	111	+and nothing else changes.
	112	+
	113	+## What it logs
	114	+
	115	+Counts and closed choices only; no path and no file text. Each row goes
	116	+through `$.telemetry.log`, so it exists only where the telemetry plugin
	117	+does:
	118	+
	119	+\| event \| when \| properties \|
	120	+\| --- \| --- \| --- \|
	121	+\| `agents_md_mode` \| once per fresh load, at `session.start` \| `mode` (`claude-md` \\| `claude-md-or-agents-md` \\| `claude-md-and-agents-md` \\| `managed-only`), `is_interactive` \|
	122	+\| `agents_md_load` \| the first context of a load, under `claude-md-or-agents-md` and `claude-md-and-agents-md` \| `mode`, `file_count` (`AGENTS.md` files handed to the engine), `import_count` (their `@` imports), `total_content_length`, `yielded` (`claude-md-or-agents-md` stood down for a `CLAUDE.md` of the project's own), `walk_failed`; with it one `$.telemetry.mark` for feature `agents_md`: `ok`, or `sad` with reason `walk_failed` \|
	123	+\| `agents_md_nested` \| a Read that attached nested files \| `mode`, `file_count` \|
	124	+
	125	+## Where it still differs from CLAUDE.md
	126	+
	127	+All of these apply only to the modes that load `AGENTS.md`,
	128	+`claude-md-or-agents-md` (the default) and `claude-md-and-agents-md`. Each
	129	+names a loader fact a plugin cannot reach through the events it has today.
	130	+
	131	+1. Nested files attach on a text `Read` only. The engine also attaches a
	132	+ directory's `CLAUDE.md` for a file `@`-mentioned in the prompt, for the
	133	+ IDE's opened file or selection, and for the `Read` tool's notebook, image
	134	+ and PDF results.
	135	+2. A nested file the plugin attaches is not registered in the loop's
	136	+ read-file state, so after a compaction the engine does not restore it
	137	+ among the recently read files (the plugin attaches it again at the next
	138	+ `Read` under that directory instead), and a change to it mid-session is
	139	+ not re-announced.
	140	+3. `/cd` carries the new tree's `CLAUDE.md` in its own notice; the plugin's
	141	+ files for the new tree arrive in the same next request through the
	142	+ engine's instructions announcement instead.
	143	+4. Paths compare by spelling; the engine resolves a symlinked alias of the
	144	+ working directory before deciding a file is inside it.
	145	+5. `--add-dir` directories contribute no `AGENTS.md`, where the engine can
	146	+ load their `CLAUDE.md`.
	147	+6. `/memory` and the `#` shortcut do not know `AGENTS.md` files, and the
	148	+ engine's own initial-load row does not count them (this plugin's
	149	+ `agents_md_load` row does).
	150	+7. An `@` import outside the working directory inside an `AGENTS.md` is
	151	+ honoured only once the approval the engine asks for a `CLAUDE.md`'s
	152	+ external imports has been given (without it the import is left out, as a
	153	+ `CLAUDE.md`'s is); the approval dialog itself is raised for `CLAUDE.md`
	154	+ imports alone.
	155	+8. A subagent that is not a fork gets a nested `AGENTS.md` at its own first
	156	+ `Read` under that directory even when its parent's loop was already given
	157	+ it; the engine does not hand such a subagent the nested `CLAUDE.md` again.
	158	+ A fork matches the engine on both sides.
	159	+
	160	+## Testing
	161	+
	162	+ claude plugin test mods/agents-md
	163	+
	164	+`tests/register.test.ts` covers the default mode: a project with `AGENTS.md`
	165	+alone gets it as a project instruction file and one transcript line naming
	166	+it, a project with a `CLAUDE.md` of its own is left to the engine without a
	167	+walk, a failed walk leaves the context as handed, and the start hands
	168	+`$.telemetry` the mode row alone where a test seats a provider for that
	169	+noun, and goes on untouched where none is seated.

		@@ -1,0 +1,28 @@
	1	+import type { InstructionFile } from 'claude-code'
	2	+
	3	+/**
	4	+ * The path of the file at the head of an instruction file's `@`-import
	5	+ * chain; the file's own path when nothing imported it.
	6	+ *
	7	+ * Follows `parent` among the given files until a file nothing imported.
	8	+ *
	9	+ * @param file the instruction file
	10	+ * @param byPath the files it may have come through, by path
	11	+ * @returns the chain head's path
	12	+ */
	13	+export function chainRootOf(
	14	+ file: InstructionFile,
	15	+ byPath: ReadonlyMap<string, InstructionFile>,
	16	+): string {
	17	+ const seen = new Set<string>([file.path])
	18	+ let path = file.path
	19	+ let parent = file.parent
	20	+
	21	+ while (parent !== undefined && !seen.has(parent)) {
	22	+ seen.add(parent)
	23	+ path = parent
	24	+ parent = byPath.get(parent)?.parent
	25	+ }
	26	+
	27	+ return path
	28	+}

		@@ -1,0 +1,12 @@
	1	+import type { InstructionFileKind } from 'claude-code'
	2	+
	3	+/**
	4	+ * The kinds `managed-only` drops: the project's checked-in and private
	5	+ * instruction files and the person's own; what it keeps is the organization's
	6	+ * and memory.
	7	+ */
	8	+export const DROPPED_KINDS: readonly InstructionFileKind[] = [
	9	+ 'project',
	10	+ 'local',
	11	+ 'user',
	12	+]

		@@ -1,0 +1,20 @@
	1	+import type { FsAncestor, InstructionFile } from 'claude-code'
	2	+
	3	+/**
	4	+ * The found AGENTS.md files as project instruction files, one per file and
	5	+ * per file it `@`-imported, in walk and load order.
	6	+ *
	7	+ * An import names the file that brought it as its parent.
	8	+ *
	9	+ * @param found what `$.fs.ancestors` found, root first
	10	+ * @returns the instruction files, kind `project`
	11	+ */
	12	+export const filesOf = (found: readonly FsAncestor[]): InstructionFile[] =>
	13	+ found.flatMap(entry =>
	14	+ entry.parts.map((part, index) => ({
	15	+ path: part.path,
	16	+ kind: 'project' as const,
	17	+ content: part.content,
	18	+ ...(index > 0 && { parent: entry.parts[0]?.path ?? part.path }),
	19	+ })),
	20	+ )

		@@ -1,0 +1,12 @@
	1	+export * from './chain-root-of.js'
	2	+export * from './dropped-kinds.js'
	3	+export * from './files-of.js'
	4	+export * from './insertion-index.js'
	5	+export * from './is-claude-file-on-walk.js'
	6	+export * from './is-kept-without-instructions.js'
	7	+export * from './is-project-own.js'
	8	+export * from './project-dir-of.js'
	9	+export * from './unseen-files.js'
	10	+export * from './with-project-files.js'
	11	+
	12	+export * as default from '.'

		@@ -1,0 +1,44 @@
	1	+import type { InstructionFile } from 'claude-code'
	2	+
	3	+import Frames from '../frames'
	4	+import { chainRootOf } from './chain-root-of.js'
	5	+import { isProjectOwn } from './is-project-own.js'
	6	+import { projectDirOf } from './project-dir-of.js'
	7	+
	8	+/**
	9	+ * Where a directory's AGENTS.md files go among the files the engine loaded,
	10	+ * as the engine orders project files root first.
	11	+ *
	12	+ * Before the first project file of a deeper directory (an imported file
	13	+ * counts with the file at the head of its chain), else after the last
	14	+ * project file, else before the engine's memory, else at the end.
	15	+ *
	16	+ * @param files the list so far
	17	+ * @param dir the directory the AGENTS.md files stand in
	18	+ * @returns the index to insert at
	19	+ */
	20	+export function insertionIndex(
	21	+ files: readonly InstructionFile[],
	22	+ dir: string,
	23	+): number {
	24	+ const byPath = new Map(files.map(file => [file.path, file]))
	25	+ const deeper = files.findIndex(
	26	+ file =>
	27	+ isProjectOwn(file) &&
	28	+ Frames.isBelow(projectDirOf(chainRootOf(file, byPath)), dir),
	29	+ )
	30	+
	31	+ if (deeper !== -1) {
	32	+ return deeper
	33	+ }
	34	+
	35	+ const lastOwn = files.findLastIndex(isProjectOwn)
	36	+
	37	+ if (lastOwn !== -1) {
	38	+ return lastOwn + 1
	39	+ }
	40	+
	41	+ const memory = files.findIndex(file => file.kind === 'memory')
	42	+
	43	+ return memory === -1 ? files.length : memory
	44	+}

		@@ -1,0 +1,39 @@
	1	+import type { InstructionFile } from 'claude-code'
	2	+
	3	+import Frames from '../frames'
	4	+import Names from '../names'
	5	+import { isProjectOwn } from './is-project-own.js'
	6	+import { projectDirOf } from './project-dir-of.js'
	7	+
	8	+/**
	9	+ * Whether a handed instruction file is a CLAUDE.md, `.claude/CLAUDE.md` or
	10	+ * CLAUDE.local.md of a directory on the walk down to the session's root.
	11	+ *
	12	+ * What makes a project "have a CLAUDE.md of its own" for
	13	+ * `claude-md-or-agents-md`, the same files the Read walk asks for; a rules
	14	+ * file, an imported file or an added directory's CLAUDE.md is not one.
	15	+ *
	16	+ * @param file the handed instruction file
	17	+ * @param root the session's project root, absolute
	18	+ * @returns true for the project's own CLAUDE.md files on the walk
	19	+ */
	20	+export function isClaudeFileOnWalk(
	21	+ file: InstructionFile,
	22	+ root: string,
	23	+): boolean {
	24	+ const isOwnClaudeFile =
	25	+ isProjectOwn(file) &&
	26	+ file.parent === undefined &&
	27	+ Names.CLAUDE_NAMES.some(name =>
	28	+ Frames.normalSpellingOf(file.path).endsWith(`/${name}`),
	29	+ )
	30	+
	31	+ if (!isOwnClaudeFile) {
	32	+ return false
	33	+ }
	34	+
	35	+ const dir = projectDirOf(file.path)
	36	+ const spelledRoot = Frames.normalSpellingOf(root)
	37	+
	38	+ return dir === spelledRoot \|\| Frames.isBelow(spelledRoot, dir)
	39	+}

		@@ -1,0 +1,13 @@
	1	+import type { InstructionFile } from 'claude-code'
	2	+
	3	+import { DROPPED_KINDS } from './dropped-kinds.js'
	4	+
	5	+/**
	6	+ * Whether `managed-only` keeps an instruction file: the organization's managed
	7	+ * files and the engine's memory stay, the project's and the person's go.
	8	+ *
	9	+ * @param file the instruction file
	10	+ * @returns true for a file `managed-only` leaves in place
	11	+ */
	12	+export const isKeptWithoutInstructions = (file: InstructionFile): boolean =>
	13	+ !DROPPED_KINDS.includes(file.kind)

		@@ -6,8 +6,324 @@
6	6	<author><name>Anthropic</name></author>
7	7	<link rel="alternate" type="text/html" href="https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md"/>
8	8	<link rel="self" type="application/atom+xml" href="https://raw.githubusercontent.com/anthropics/claude-code/main/feed.xml"/>
9		- <updated>2026-09-15T20:22:55Z</updated>
	9	+ <updated>2026-09-18T18:06:17Z</updated>
10	10	<entry>
	11	+ <id>https://github.com/anthropics/claude-code/releases/tag/v2.1.277</id>
	12	+ <title>Claude Code v2.1.277</title>
	13	+ <link rel="alternate" type="text/html" href="https://github.com/anthropics/claude-code/releases/tag/v2.1.277"/>
	14	+ <updated>2026-09-18T18:06:17Z</updated>
	15	+ <content type="html"><p>• Added AGENTS.md support: in a project with no CLAUDE.md, Claude Code reads AGENTS.md instead; change it under "Project instructions" in /config (not yet on Bedrock, Vertex or Foundry)</p>
	16	+<p>• Added CLAUDE_GATEWAY_PROXY_IS_EGRESS_BOUNDARY=1 for Claude apps gateways whose only egress is a forward proxy: every outbound request hands the proxy the hostname instead of resolving it locally</p>
	17	+<p>• Added an optional headers: map on Claude apps gateway upstreams, to send static headers to a proxy you run in front of a provider</p>
	18	+<p>• Added a line saying a background task's update is waiting when it finishes while a panel such as /tasks is open</p>
	19	+<p>• Fixed claude -p and Agent SDK sessions that could hang with no result after an internal error; they now report the error and exit with code 1</p>
	20	+<p>• Fixed conversations failing every request with "text content blocks must be non-empty" when an earlier assistant turn held an empty text block beside other content, including after --resume</p>
	21	+<p>• Fixed being unexpectedly logged out when an older Claude Code build (for example an IDE extension's bundled CLI) runs on the same machine as the current one</p>
	22	+<p>• Fixed interactive start-up hanging or showing an error for ANTHROPIC_API_KEY users when ~/.claude.json holds a malformed customApiKeyResponses value</p>
	23	+<p>• Fixed update checks erroring every 30 minutes, and claude update hanging when a minimum or maximum version is set, if a proxy returns an invalid version; a malformed minimumVersion is now ignored</p>
	24	+<p>• Fixed claude update on winget- or apk-managed installs reporting "up to date" when the version lookup failed</p>
	25	+<p>• Fixed claude plugin install sometimes failing and breaking the installed copy when reinstalling a plugin version that a session or another program was using; an unchanged copy is now left alone</p>
	26	+<p>• Fixed Grep and Glob reporting no matches when the search could not start because the system was out of processes, memory or file handles; they now return an error saying so</p>
	27	+<p>• Fixed the Write tool silently ending the turn as a declined permission when the target path is an existing directory; it now reports a clear error</p>
	28	+<p>• Fixed the Edit tool treating an escaped backslash followed by uXXXX text as a \uXXXX escape, which could make an edit of a non-ASCII character rewrite an escaped backslash sequence instead</p>
	29	+<p>• Fixed the Edit tool reporting "Invalid regular expression: regular expression too large" instead of "String not found in file" when a very large edit containing non-ASCII text did not match the file</p>
	30	+<p>• Fixed a turn ending early with "Path contains null bytes" when a tool call's file path contained \u0000 written as an escape sequence; escaped control characters now stay as literal text</p>
	31	+<p>• Fixed background sessions (claude --bg) exiting when a plugin's LSP server exited or closed its stdin</p>
	32	+<p>• Fixed a crash ("Type error") when opening /mcp or /plugin manage with a malformed claudeAiMcpEverConnected value in ~/.claude.json</p>
	33	+<p>• Fixed a crash at launch when ~/.claude.json holds a malformed theme value</p>
	34	+<p>• Fixed a crash ("unrecoverable interface error") when the prompt held text containing terminal color codes, for example a prompt recalled from history or text loaded from the external editor</p>
	35	+<p>• Fixed a crash when resuming a session whose saved history holds an assistant message stored as a plain string</p>
	36	+<p>• Fixed sessions on slow or heavily loaded machines sometimes exiting with "Claude Code exited after an unrecoverable interface error" when the first spinner appeared</p>
	37	+<p>• Fixed a rare case where the screen could stop updating for the rest of the session after an internal rendering error</p>
	38	+<p>• Fixed a rare case on Windows where a turn could stop with an error such as "Out of memory" right after Claude replied, so that reply's tool calls never ran</p>
	39	+<p>• Fixed sessions continued after /clear (restart, --continue, --resume) missing part of their first message when a SessionStart hook printed output, causing a full prompt-cache miss</p>
	40	+<p>• Fixed messages from other agents (such as a subagent's SendMessage) that arrived mid-turn showing up below the "Ran N shell commands" row instead of where they arrived</p>
	41	+<p>• Fixed the "copied" notice not appearing after drag-selecting text in the fullscreen /resume picker and other panels that cover the prompt area</p>
	42	+<p>• Fixed $TMPDIR expanding empty in Bash commands that run outside the sandbox while sandboxing is enabled</p>
	43	+<p>• Fixed WebFetch and WebSearch in Cowork cloud sessions not telling Claude why a request was refused, such as a used-up fetch budget or an admin policy</p>

		@@ -1,0 +1,11 @@
	1	+import type { Mode } from './types'
	2	+
	3	+/**
	4	+ * Every mode, in the order the option's description names them.
	5	+ */
	6	+export const MODES: readonly Mode[] = [
	7	+ 'claude-md',
	8	+ 'claude-md-or-agents-md',
	9	+ 'claude-md-and-agents-md',
	10	+ 'managed-only',
	11	+]

		@@ -1,0 +1,304 @@
	1	+import type {
	2	+ EngineInterface,
	3	+ FsAncestor,
	4	+ InstructionFile,
	5	+ On,
	6	+ PluginOptions,
	7	+} from 'claude-code'
	8	+
	9	+import Files from './files'
	10	+import Frames from './frames'
	11	+import Modes from './modes'
	12	+import Names from './names'
	13	+import Switches from './switches'
	14	+import Telemetry from './telemetry'
	15	+
	16	+/**
	17	+ * No files: what a walk not taken, or one that failed, found.
	18	+ */
	19	+const NONE: readonly FsAncestor[] = []
	20	+
	21	+/**
	22	+ * Registers the plugin's hooks for the mode `instructionFiles` names
	23	+ * (`claude-md-or-agents-md` when unset; the manifest lists the four, nothing
	24	+ * else arrives).
	25	+ *
	26	+ * `claude-md`: nothing beyond the usage row, the engine's CLAUDE.md walk
	27	+ * standing alone. `managed-only`: the project's and the person's instruction
	28	+ * files dropped, the organization's kept. `claude-md-or-agents-md` (a project
	29	+ * with none of its own) and `claude-md-and-agents-md`: AGENTS.md files joined
	30	+ * to the engine's instruction files, nested ones on a Read except in a run
	31	+ * where the engine attaches nothing to a turn (--bare, which sets
	32	+ * CLAUDE_CODE_SIMPLE, or CLAUDE_CODE_DISABLE_ATTACHMENTS). Every mode sends
	33	+ * its usage rows through `$.telemetry` where that noun is seated and drops
	34	+ * them where it is not.
	35	+ *
	36	+ * @param on the engine's registrar
	37	+ * @param options the plugin's options; `instructionFiles` is `claude-md`,
	38	+ * `claude-md-or-agents-md`, `claude-md-and-agents-md` or `managed-only`
	39	+ */

		@@ -1,0 +1,16 @@
	1	+import Frames from '../frames'
	2	+
	3	+/**
	4	+ * The project directory an instruction file speaks for: the directory
	5	+ * holding it, or the one holding its `.claude` folder.
	6	+ *
	7	+ * @param path the file's path, absolute
	8	+ * @returns the directory, in the walk's normal spelling
	9	+ */
	10	+export function projectDirOf(path: string): string {
	11	+ const spelled = Frames.normalSpellingOf(path)
	12	+ const claudeAt = spelled.lastIndexOf('/.claude/')
	13	+ const cut = claudeAt === -1 ? spelled.lastIndexOf('/') : claudeAt
	14	+
	15	+ return cut <= 0 ? '/' : spelled.slice(0, cut)
	16	+}

		@@ -1,0 +1,40 @@
	1	+import type { InstructionFile } from 'claude-code'
	2	+
	3	+import Frames from '../frames'
	4	+import { isProjectOwn } from './is-project-own.js'
	5	+
	6	+/**
	7	+ * The candidate files the engine has not already loaded, and no candidate
	8	+ * twice.
	9	+ *
	10	+ * Leaves out one whose path is among the handed files (a CLAUDE.md that
	11	+ * `@`-imports AGENTS.md) or whose text equals a handed project file's (a
	12	+ * CLAUDE.md symlinked to AGENTS.md); other kinds' text does not count.
	13	+ *
	14	+ * @param candidates the AGENTS.md instruction files of a walk
	15	+ * @param handed the files the engine handed the hook
	16	+ * @returns the candidates worth adding, in order
	17	+ */
	18	+export function unseenFiles(
	19	+ candidates: readonly InstructionFile[],
	20	+ handed: readonly InstructionFile[],
	21	+): InstructionFile[] {
	22	+ const paths = new Set(handed.map(file => Frames.normalSpellingOf(file.path)))
	23	+ const texts = new Set(
	24	+ handed.filter(isProjectOwn).map(file => file.content.trim()),
	25	+ )
	26	+ const kept: InstructionFile[] = []
	27	+
	28	+ for (const file of candidates) {
	29	+ const path = Frames.normalSpellingOf(file.path)
	30	+ const text = file.content.trim()
	31	+ const isSeen = paths.has(path) \|\| (text !== '' && texts.has(text))
	32	+
	33	+ if (!isSeen) {
	34	+ paths.add(path)
	35	+ kept.push(file)
	36	+ }
	37	+ }
	38	+
	39	+ return kept
	40	+}

		@@ -1,0 +1,37 @@
	1	+import type { InstructionFile } from 'claude-code'
	2	+
	3	+import { chainRootOf } from './chain-root-of.js'
	4	+import { insertionIndex } from './insertion-index.js'
	5	+import { projectDirOf } from './project-dir-of.js'
	6	+
	7	+/**
	8	+ * The engine's instruction files with the AGENTS.md files placed where the
	9	+ * engine would place a project file of their directory; as is when none.
	10	+ *
	11	+ * @param handed the files the engine handed the hook, in its order
	12	+ * @param added the AGENTS.md instruction files to add, root first
	13	+ * @returns the merged list
	14	+ */
	15	+export function withProjectFiles(
	16	+ handed: readonly InstructionFile[],
	17	+ added: readonly InstructionFile[],
	18	+): readonly InstructionFile[] {
	19	+ if (added.length === 0) {
	20	+ return handed
	21	+ }
	22	+
	23	+ const merged = [...handed]
	24	+ const byPath = new Map(added.map(file => [file.path, file]))
	25	+ const byDir = new Map<string, InstructionFile[]>()
	26	+
	27	+ for (const file of added) {
	28	+ const dir = projectDirOf(chainRootOf(file, byPath))
	29	+ byDir.set(dir, [...(byDir.get(dir) ?? []), file])
	30	+ }
	31	+
	32	+ for (const [dir, files] of byDir) {
	33	+ merged.splice(insertionIndex(merged, dir), 0, ...files)
	34	+ }
	35	+
	36	+ return merged
	37	+}

		@@ -1,0 +1,33 @@
	1	+/**
	2	+ * A path the model handed a Read, made absolute as the Read tool reads it.
	3	+ *
	4	+ * As given when it is absolute; under the home directory (less any trailing
	5	+ * separator) for `~` or a `~/` spelling when a home is known; else under the
	6	+ * working directory.
	7	+ *
	8	+ * @param path the Read's `file_path` as the model spelled it
	9	+ * @param root the session's working directory, absolute
	10	+ * @param home the home directory, absolute, when the environment names one
	11	+ * @returns the absolute path
	12	+ */
	13	+export function absoluteOf(path: string, root: string, home?: string): string {
	14	+ const isHomeSpelling = path === '~' \|\| path.startsWith('~/')
	15	+
	16	+ if (isHomeSpelling && home !== undefined) {
	17	+ const bare = home.replace(/(?<=.)[\\/]+$/, '')
	18	+
	19	+ return path === '~' ? bare : `${bare}${separatorOf(bare)}${path.slice(2)}`
	20	+ }
	21	+
	22	+ return /^(?:[A-Za-z]:)?[\\/]/.test(path)
	23	+ ? path
	24	+ : `${root}${separatorOf(root)}${path}`
	25	+}
	26	+
	27	+/**
	28	+ * The separator a directory's own spelling uses.
	29	+ *
	30	+ * @param dir an absolute directory
	31	+ * @returns `\` for a backslash spelling, `/` otherwise
	32	+ */
	33	+const separatorOf = (dir: string): string => (dir.includes('\\') ? '\\' : '/')

		@@ -1,0 +1,8 @@
	1	+export * from './absolute-of.js'
	2	+export * from './is-below.js'
	3	+export * from './is-file-at.js'
	4	+export * from './nested-frame.js'
	5	+export * from './normal-spelling-of.js'
	6	+export * from './outside-claude-dirs.js'
	7	+
	8	+export * as default from '.'

		@@ -1,0 +1,12 @@
	1	+import { normalSpellingOf } from './normal-spelling-of.js'
	2	+
	3	+/**
	4	+ * Whether a path lies strictly inside a directory, by spelling (no links
	5	+ * followed), as the engine bounds its nested CLAUDE.md walk.
	6	+ *
	7	+ * @param path an absolute path
	8	+ * @param dir an absolute directory
	9	+ * @returns true when `path` is under `dir` and not `dir` itself
	10	+ */
	11	+export const isBelow = (path: string, dir: string): boolean =>
	12	+ normalSpellingOf(path).startsWith(`${normalSpellingOf(dir)}/`)

		@@ -1,0 +1,14 @@
	1	+import type { InstructionFile } from 'claude-code'
	2	+
	3	+import { normalSpellingOf } from './normal-spelling-of.js'
	4	+
	5	+/**
	6	+ * Whether an instruction file is the file at a path, however either is
	7	+ * spelled.
	8	+ *
	9	+ * @param file the instruction file
	10	+ * @param path an absolute path
	11	+ * @returns true when both name one path
	12	+ */
	13	+export const isFileAt = (file: InstructionFile, path: string): boolean =>
	14	+ normalSpellingOf(file.path) === normalSpellingOf(path)

		@@ -1,0 +1,9 @@
	1	+/**
	2	+ * A path with one separator spelling and no trailing separator, so two
	3	+ * spellings of one path compare equal (`C:/repo/x` and `C:\repo\x`).
	4	+ *
	5	+ * @param path an absolute path as a walk or the model spelled it
	6	+ * @returns the path with `/` for every separator, no trailing one
	7	+ */
	8	+export const normalSpellingOf = (path: string): string =>
	9	+ path.replaceAll('\\', '/').replace(/(?<=.)\/+$/, '')

		@@ -1,0 +1,21 @@
	1	+import type { FsAncestor } from 'claude-code'
	2	+
	3	+/**
	4	+ * The AGENTS.md files standing in a directory with no CLAUDE.md of its own,
	5	+ * in walk order.
	6	+ *
	7	+ * Where the engine attaches a directory's CLAUDE.md, that directory's
	8	+ * AGENTS.md is left out rather than merged.
	9	+ *
	10	+ * @param agents the AGENTS.md files of a walk, root first
	11	+ * @param claude the CLAUDE.md files of the same walk
	12	+ * @returns the AGENTS.md files whose directory holds no CLAUDE.md
	13	+ */
	14	+export function outsideClaudeDirs(
	15	+ agents: readonly FsAncestor[],
	16	+ claude: readonly FsAncestor[],
	17	+): readonly FsAncestor[] {
	18	+ const taken = new Set(claude.map(file => file.dir))
	19	+
	20	+ return agents.filter(file => !taken.has(file.dir))
	21	+}

		@@ -1,0 +1,4 @@
	1	+{
	2	+ "description": "AGENTS.md up and down the tree like CLAUDE.md, by the instructionFiles option: under claude-md-or-agents-md (the default; where the project has no CLAUDE.md of its own on the walk) or claude-md-and-agents-md, a prompt.context hook hands every ancestor AGENTS.md to the engine among its instruction files, skipping ones it already loaded, and a tool.call hook on Read attaches the AGENTS.md files of the directories between the project root and a read file, once a conversation and agent loop (an Agent-tool fork inherits its parent's), except in a run where the engine attaches nothing to a turn (--bare, CLAUDE_CODE_DISABLE_ATTACHMENTS); under managed-only, a prompt.context hook keeps only the organization's managed files and memory; under claude-md, nothing is added; in every mode a session.start hook sends the plugin's usage row through $.telemetry where that noun is seated",
	3	+ "modules": ["./register.ts"]
	4	+}

		@@ -1,0 +1,9 @@
	1	+export * from './files'
	2	+export * from './frames'
	3	+export * from './modes'
	4	+export * from './names'
	5	+export * from './register.js'
	6	+export * from './switches'
	7	+export * from './telemetry'
	8	+
	9	+export * as default from '.'

		@@ -1,0 +1,7 @@
	1	+export * from './default-mode.js'
	2	+export * from './legacy-mode-of.js'
	3	+export * from './mode-of.js'
	4	+export * from './modes.js'
	5	+export * from './types'
	6	+
	7	+export * as default from '.'

		@@ -1,0 +1,13 @@
	1	+import { DEFAULT_MODE } from './default-mode.js'
	2	+import { MODES } from './modes.js'
	3	+import type { Mode } from './types'
	4	+
	5	+/**
	6	+ * The mode a stored option names; unset, a typo or a value of another type
	7	+ * names the default.
	8	+ *
	9	+ * @param value the `instructionFiles` option as stored
	10	+ * @returns the mode named, DEFAULT_MODE for anything else
	11	+ */
	12	+export const modeOf = (value: unknown): Mode =>
	13	+ MODES.find(mode => mode === value) ?? DEFAULT_MODE

		@@ -1,0 +1,3 @@
	1	+export * from './mode'
	2	+
	3	+export * as default from '.'

		@@ -1,0 +1,3 @@
	1	+export type * from './mode.js'
	2	+
	3	+export * as default from '.'

		@@ -1,0 +1,15 @@
	1	+/**
	2	+ * Which project instructions the plugin loads: the `instructionFiles`
	3	+ * option.
	4	+ *
	5	+ * `claude-md` loads none of its own, the engine's CLAUDE.md walk standing
	6	+ * alone; `claude-md-or-agents-md` loads AGENTS.md where the project has no
	7	+ * CLAUDE.md; `claude-md-and-agents-md` loads AGENTS.md beside CLAUDE.md;
	8	+ * `managed-only` loads neither, taking the project's and the person's
	9	+ * CLAUDE.md files out of the conversation's context.
	10	+ */
	11	+export type Mode =
	12	+ \| 'claude-md'
	13	+ \| 'claude-md-or-agents-md'
	14	+ \| 'claude-md-and-agents-md'
	15	+ \| 'managed-only'

		@@ -1,0 +1,5 @@
	1	+/**
	2	+ * The instruction files read in each directory, in the order the engine
	3	+ * reads CLAUDE.md and .claude/CLAUDE.md.
	4	+ */
	5	+export const AGENTS_NAMES = ['AGENTS.md', '.claude/AGENTS.md'] as const

		@@ -1,0 +1,11 @@
	1	+/**
	2	+ * The engine's own instruction files, as it reads them in each directory of
	3	+ * its walk and under a read file.
	4	+ *
	5	+ * A directory holding one has project instructions the engine loads itself.
	6	+ */
	7	+export const CLAUDE_NAMES = [
	8	+ 'CLAUDE.md',
	9	+ '.claude/CLAUDE.md',
	10	+ 'CLAUDE.local.md',
	11	+] as const

		@@ -1,0 +1,5 @@
	1	+export * from './agents-names.js'
	2	+export * from './claude-names.js'
	3	+export * from './main-loop.js'
	4	+
	5	+export * as default from '.'

		@@ -1,0 +1,5 @@
	1	+/**
	2	+ * The key the main conversation's delivered files are kept under; a
	3	+ * subagent's are kept under its `agentId`.
	4	+ */
	5	+export const MAIN_LOOP = 'main'