forked from sanbuphy/learn-coding-agent
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathclaudemd.ts
More file actions
1479 lines (1344 loc) · 45.3 KB
/
claudemd.ts
File metadata and controls
1479 lines (1344 loc) · 45.3 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/**
* Files are loaded in the following order:
*
* 1. Managed memory (eg. /etc/claude-code/CLAUDE.md) - Global instructions for all users
* 2. User memory (~/.claude/CLAUDE.md) - Private global instructions for all projects
* 3. Project memory (CLAUDE.md, .claude/CLAUDE.md, and .claude/rules/*.md in project roots) - Instructions checked into the codebase
* 4. Local memory (CLAUDE.local.md in project roots) - Private project-specific instructions
*
* Files are loaded in reverse order of priority, i.e. the latest files are highest priority
* with the model paying more attention to them.
*
* File discovery:
* - User memory is loaded from the user's home directory
* - Project and Local files are discovered by traversing from the current directory up to root
* - Files closer to the current directory have higher priority (loaded later)
* - CLAUDE.md, .claude/CLAUDE.md, and all .md files in .claude/rules/ are checked in each directory for Project memory
*
* Memory @include directive:
* - Memory files can include other files using @ notation
* - Syntax: @path, @./relative/path, @~/home/path, or @/absolute/path
* - @path (without prefix) is treated as a relative path (same as @./path)
* - Works in leaf text nodes only (not inside code blocks or code strings)
* - Included files are added as separate entries before the including file
* - Circular references are prevented by tracking processed files
* - Non-existent files are silently ignored
*/
import { feature } from 'bun:bundle'
import ignore from 'ignore'
import memoize from 'lodash-es/memoize.js'
import { Lexer } from 'marked'
import {
basename,
dirname,
extname,
isAbsolute,
join,
parse,
relative,
sep,
} from 'path'
import picomatch from 'picomatch'
import { logEvent } from 'src/services/analytics/index.js'
import {
getAdditionalDirectoriesForClaudeMd,
getOriginalCwd,
} from '../bootstrap/state.js'
import { truncateEntrypointContent } from '../memdir/memdir.js'
import { getAutoMemEntrypoint, isAutoMemoryEnabled } from '../memdir/paths.js'
import { getFeatureValue_CACHED_MAY_BE_STALE } from '../services/analytics/growthbook.js'
import {
getCurrentProjectConfig,
getManagedClaudeRulesDir,
getMemoryPath,
getUserClaudeRulesDir,
} from './config.js'
import { logForDebugging } from './debug.js'
import { logForDiagnosticsNoPII } from './diagLogs.js'
import { getClaudeConfigHomeDir, isEnvTruthy } from './envUtils.js'
import { getErrnoCode } from './errors.js'
import { normalizePathForComparison } from './file.js'
import { cacheKeys, type FileStateCache } from './fileStateCache.js'
import {
parseFrontmatter,
splitPathInFrontmatter,
} from './frontmatterParser.js'
import { getFsImplementation, safeResolvePath } from './fsOperations.js'
import { findCanonicalGitRoot, findGitRoot } from './git.js'
import {
executeInstructionsLoadedHooks,
hasInstructionsLoadedHook,
type InstructionsLoadReason,
type InstructionsMemoryType,
} from './hooks.js'
import type { MemoryType } from './memory/types.js'
import { expandPath } from './path.js'
import { pathInWorkingPath } from './permissions/filesystem.js'
import { isSettingSourceEnabled } from './settings/constants.js'
import { getInitialSettings } from './settings/settings.js'
/* eslint-disable @typescript-eslint/no-require-imports */
const teamMemPaths = feature('TEAMMEM')
? (require('../memdir/teamMemPaths.js') as typeof import('../memdir/teamMemPaths.js'))
: null
/* eslint-enable @typescript-eslint/no-require-imports */
let hasLoggedInitialLoad = false
const MEMORY_INSTRUCTION_PROMPT =
'Codebase and user instructions are shown below. Be sure to adhere to these instructions. IMPORTANT: These instructions OVERRIDE any default behavior and you MUST follow them exactly as written.'
// Recommended max character count for a memory file
export const MAX_MEMORY_CHARACTER_COUNT = 40000
// File extensions that are allowed for @include directives
// This prevents binary files (images, PDFs, etc.) from being loaded into memory
const TEXT_FILE_EXTENSIONS = new Set([
// Markdown and text
'.md',
'.txt',
'.text',
// Data formats
'.json',
'.yaml',
'.yml',
'.toml',
'.xml',
'.csv',
// Web
'.html',
'.htm',
'.css',
'.scss',
'.sass',
'.less',
// JavaScript/TypeScript
'.js',
'.ts',
'.tsx',
'.jsx',
'.mjs',
'.cjs',
'.mts',
'.cts',
// Python
'.py',
'.pyi',
'.pyw',
// Ruby
'.rb',
'.erb',
'.rake',
// Go
'.go',
// Rust
'.rs',
// Java/Kotlin/Scala
'.java',
'.kt',
'.kts',
'.scala',
// C/C++
'.c',
'.cpp',
'.cc',
'.cxx',
'.h',
'.hpp',
'.hxx',
// C#
'.cs',
// Swift
'.swift',
// Shell
'.sh',
'.bash',
'.zsh',
'.fish',
'.ps1',
'.bat',
'.cmd',
// Config
'.env',
'.ini',
'.cfg',
'.conf',
'.config',
'.properties',
// Database
'.sql',
'.graphql',
'.gql',
// Protocol
'.proto',
// Frontend frameworks
'.vue',
'.svelte',
'.astro',
// Templating
'.ejs',
'.hbs',
'.pug',
'.jade',
// Other languages
'.php',
'.pl',
'.pm',
'.lua',
'.r',
'.R',
'.dart',
'.ex',
'.exs',
'.erl',
'.hrl',
'.clj',
'.cljs',
'.cljc',
'.edn',
'.hs',
'.lhs',
'.elm',
'.ml',
'.mli',
'.f',
'.f90',
'.f95',
'.for',
// Build files
'.cmake',
'.make',
'.makefile',
'.gradle',
'.sbt',
// Documentation
'.rst',
'.adoc',
'.asciidoc',
'.org',
'.tex',
'.latex',
// Lock files (often text-based)
'.lock',
// Misc
'.log',
'.diff',
'.patch',
])
export type MemoryFileInfo = {
path: string
type: MemoryType
content: string
parent?: string // Path of the file that included this one
globs?: string[] // Glob patterns for file paths this rule applies to
// True when auto-injection transformed `content` (stripped HTML comments,
// stripped frontmatter, truncated MEMORY.md) such that it no longer matches
// the bytes on disk. When set, `rawContent` holds the unmodified disk bytes
// so callers can cache a `isPartialView` readFileState entry — presence in
// cache provides dedup + change detection, but Edit/Write still require an
// explicit Read before proceeding.
contentDiffersFromDisk?: boolean
rawContent?: string
}
function pathInOriginalCwd(path: string): boolean {
return pathInWorkingPath(path, getOriginalCwd())
}
/**
* Parses raw content to extract both content and glob patterns from frontmatter
* @param rawContent Raw file content with frontmatter
* @returns Object with content and globs (undefined if no paths or match-all pattern)
*/
function parseFrontmatterPaths(rawContent: string): {
content: string
paths?: string[]
} {
const { frontmatter, content } = parseFrontmatter(rawContent)
if (!frontmatter.paths) {
return { content }
}
const patterns = splitPathInFrontmatter(frontmatter.paths)
.map(pattern => {
// Remove /** suffix - ignore library treats 'path' as matching both
// the path itself and everything inside it
return pattern.endsWith('/**') ? pattern.slice(0, -3) : pattern
})
.filter((p: string) => p.length > 0)
// If all patterns are ** (match-all), treat as no globs (undefined)
// This means the file applies to all paths
if (patterns.length === 0 || patterns.every((p: string) => p === '**')) {
return { content }
}
return { content, paths: patterns }
}
/**
* Strip block-level HTML comments (<!-- ... -->) from markdown content.
*
* Uses the marked lexer to identify comments at the block level only, so
* comments inside inline code spans and fenced code blocks are preserved.
* Inline HTML comments inside a paragraph are also left intact; the intended
* use case is authorial notes that occupy their own lines.
*
* Unclosed comments (`<!--` with no matching `-->`) are left in place so a
* typo doesn't silently swallow the rest of the file.
*/
export function stripHtmlComments(content: string): {
content: string
stripped: boolean
} {
if (!content.includes('<!--')) {
return { content, stripped: false }
}
// gfm:false is fine here — html-block detection is a CommonMark rule.
return stripHtmlCommentsFromTokens(new Lexer({ gfm: false }).lex(content))
}
function stripHtmlCommentsFromTokens(tokens: ReturnType<Lexer['lex']>): {
content: string
stripped: boolean
} {
let result = ''
let stripped = false
// A well-formed HTML comment span. Non-greedy so multiple comments on the
// same line are matched independently; [\s\S] to span newlines.
const commentSpan = /<!--[\s\S]*?-->/g
for (const token of tokens) {
if (token.type === 'html') {
const trimmed = token.raw.trimStart()
if (trimmed.startsWith('<!--') && trimmed.includes('-->')) {
// Per CommonMark, a type-2 HTML block ends at the *line* containing
// `-->`, so text after `-->` on that line is part of this token.
// Strip only the comment spans and keep any residual content.
const residue = token.raw.replace(commentSpan, '')
stripped = true
if (residue.trim().length > 0) {
// Residual content exists (e.g. `<!-- note --> Use bun`): keep it.
result += residue
}
continue
}
}
result += token.raw
}
return { content: result, stripped }
}
/**
* Parses raw memory file content into a MemoryFileInfo. Pure function — no I/O.
*
* When includeBasePath is given, @include paths are resolved in the same lex
* pass and returned alongside the parsed file (so processMemoryFile doesn't
* need to lex the same content a second time).
*/
function parseMemoryFileContent(
rawContent: string,
filePath: string,
type: MemoryType,
includeBasePath?: string,
): { info: MemoryFileInfo | null; includePaths: string[] } {
// Skip non-text files to prevent loading binary data (images, PDFs, etc.) into memory
const ext = extname(filePath).toLowerCase()
if (ext && !TEXT_FILE_EXTENSIONS.has(ext)) {
logForDebugging(`Skipping non-text file in @include: ${filePath}`)
return { info: null, includePaths: [] }
}
const { content: withoutFrontmatter, paths } =
parseFrontmatterPaths(rawContent)
// Lex once so strip and @include-extract share the same tokens. gfm:false
// is required by extract (so ~/path doesn't tokenize as strikethrough) and
// doesn't affect strip (html blocks are a CommonMark rule).
const hasComment = withoutFrontmatter.includes('<!--')
const tokens =
hasComment || includeBasePath !== undefined
? new Lexer({ gfm: false }).lex(withoutFrontmatter)
: undefined
// Only rebuild via tokens when a comment actually needs stripping —
// marked normalises \r\n during lex, so round-tripping a CRLF file
// through token.raw would spuriously flip contentDiffersFromDisk.
const strippedContent =
hasComment && tokens
? stripHtmlCommentsFromTokens(tokens).content
: withoutFrontmatter
const includePaths =
tokens && includeBasePath !== undefined
? extractIncludePathsFromTokens(tokens, includeBasePath)
: []
// Truncate MEMORY.md entrypoints to the line AND byte caps
let finalContent = strippedContent
if (type === 'AutoMem' || type === 'TeamMem') {
finalContent = truncateEntrypointContent(strippedContent).content
}
// Covers frontmatter strip, HTML comment strip, and MEMORY.md truncation
const contentDiffersFromDisk = finalContent !== rawContent
return {
info: {
path: filePath,
type,
content: finalContent,
globs: paths,
contentDiffersFromDisk,
rawContent: contentDiffersFromDisk ? rawContent : undefined,
},
includePaths,
}
}
function handleMemoryFileReadError(error: unknown, filePath: string): void {
const code = getErrnoCode(error)
// ENOENT = file doesn't exist, EISDIR = is a directory — both expected
if (code === 'ENOENT' || code === 'EISDIR') {
return
}
// Log permission errors (EACCES) as they're actionable
if (code === 'EACCES') {
// Don't log the full file path to avoid PII/security issues
logEvent('tengu_claude_md_permission_error', {
is_access_error: 1,
has_home_dir: filePath.includes(getClaudeConfigHomeDir()) ? 1 : 0,
})
}
}
/**
* Used by processMemoryFile → getMemoryFiles so the event loop stays
* responsive during the directory walk (many readFile attempts, most
* ENOENT). When includeBasePath is given, @include paths are resolved in
* the same lex pass and returned alongside the parsed file.
*/
async function safelyReadMemoryFileAsync(
filePath: string,
type: MemoryType,
includeBasePath?: string,
): Promise<{ info: MemoryFileInfo | null; includePaths: string[] }> {
try {
const fs = getFsImplementation()
const rawContent = await fs.readFile(filePath, { encoding: 'utf-8' })
return parseMemoryFileContent(rawContent, filePath, type, includeBasePath)
} catch (error) {
handleMemoryFileReadError(error, filePath)
return { info: null, includePaths: [] }
}
}
type MarkdownToken = {
type: string
text?: string
href?: string
tokens?: MarkdownToken[]
raw?: string
items?: MarkdownToken[]
}
// Extract @path include references from pre-lexed tokens and resolve to
// absolute paths. Skips html tokens so @paths inside block comments are
// ignored — the caller may pass pre-strip tokens.
function extractIncludePathsFromTokens(
tokens: ReturnType<Lexer['lex']>,
basePath: string,
): string[] {
const absolutePaths = new Set<string>()
// Extract @paths from a text string and add resolved paths to absolutePaths.
function extractPathsFromText(textContent: string) {
const includeRegex = /(?:^|\s)@((?:[^\s\\]|\\ )+)/g
let match
while ((match = includeRegex.exec(textContent)) !== null) {
let path = match[1]
if (!path) continue
// Strip fragment identifiers (#heading, #section-name, etc.)
const hashIndex = path.indexOf('#')
if (hashIndex !== -1) {
path = path.substring(0, hashIndex)
}
if (!path) continue
// Unescape the spaces in the path
path = path.replace(/\\ /g, ' ')
// Accept @path, @./path, @~/path, or @/path
if (path) {
const isValidPath =
path.startsWith('./') ||
path.startsWith('~/') ||
(path.startsWith('/') && path !== '/') ||
(!path.startsWith('@') &&
!path.match(/^[#%^&*()]+/) &&
path.match(/^[a-zA-Z0-9._-]/))
if (isValidPath) {
const resolvedPath = expandPath(path, dirname(basePath))
absolutePaths.add(resolvedPath)
}
}
}
}
// Recursively process elements to find text nodes
function processElements(elements: MarkdownToken[]) {
for (const element of elements) {
if (element.type === 'code' || element.type === 'codespan') {
continue
}
// For html tokens that contain comments, strip the comment spans and
// check the residual for @paths (e.g. `<!-- note --> @./file.md`).
// Other html tokens (non-comment tags) are skipped entirely.
if (element.type === 'html') {
const raw = element.raw || ''
const trimmed = raw.trimStart()
if (trimmed.startsWith('<!--') && trimmed.includes('-->')) {
const commentSpan = /<!--[\s\S]*?-->/g
const residue = raw.replace(commentSpan, '')
if (residue.trim().length > 0) {
extractPathsFromText(residue)
}
}
continue
}
// Process text nodes
if (element.type === 'text') {
extractPathsFromText(element.text || '')
}
// Recurse into children tokens
if (element.tokens) {
processElements(element.tokens)
}
// Special handling for list structures
if (element.items) {
processElements(element.items)
}
}
}
processElements(tokens as MarkdownToken[])
return [...absolutePaths]
}
const MAX_INCLUDE_DEPTH = 5
/**
* Checks whether a CLAUDE.md file path is excluded by the claudeMdExcludes setting.
* Only applies to User, Project, and Local memory types.
* Managed, AutoMem, and TeamMem types are never excluded.
*
* Matches both the original path and the realpath-resolved path to handle symlinks
* (e.g., /tmp -> /private/tmp on macOS).
*/
function isClaudeMdExcluded(filePath: string, type: MemoryType): boolean {
if (type !== 'User' && type !== 'Project' && type !== 'Local') {
return false
}
const patterns = getInitialSettings().claudeMdExcludes
if (!patterns || patterns.length === 0) {
return false
}
const matchOpts = { dot: true }
const normalizedPath = filePath.replaceAll('\\', '/')
// Build an expanded pattern list that includes realpath-resolved versions of
// absolute patterns. This handles symlinks like /tmp -> /private/tmp on macOS:
// the user writes "/tmp/project/CLAUDE.md" in their exclude, but the system
// resolves the CWD to "/private/tmp/project/...", so the file path uses the
// real path. By resolving the patterns too, both sides match.
const expandedPatterns = resolveExcludePatterns(patterns).filter(
p => p.length > 0,
)
if (expandedPatterns.length === 0) {
return false
}
return picomatch.isMatch(normalizedPath, expandedPatterns, matchOpts)
}
/**
* Expands exclude patterns by resolving symlinks in absolute path prefixes.
* For each absolute pattern (starting with /), tries to resolve the longest
* existing directory prefix via realpathSync and adds the resolved version.
* Glob patterns (containing *) have their static prefix resolved.
*/
function resolveExcludePatterns(patterns: string[]): string[] {
const fs = getFsImplementation()
const expanded: string[] = patterns.map(p => p.replaceAll('\\', '/'))
for (const normalized of expanded) {
// Only resolve absolute patterns — glob-only patterns like "**/*.md" don't have
// a filesystem prefix to resolve
if (!normalized.startsWith('/')) {
continue
}
// Find the static prefix before any glob characters
const globStart = normalized.search(/[*?{[]/)
const staticPrefix =
globStart === -1 ? normalized : normalized.slice(0, globStart)
const dirToResolve = dirname(staticPrefix)
try {
// sync IO: called from sync context (isClaudeMdExcluded -> processMemoryFile -> getMemoryFiles)
const resolvedDir = fs.realpathSync(dirToResolve).replaceAll('\\', '/')
if (resolvedDir !== dirToResolve) {
const resolvedPattern =
resolvedDir + normalized.slice(dirToResolve.length)
expanded.push(resolvedPattern)
}
} catch {
// Directory doesn't exist; skip resolution for this pattern
}
}
return expanded
}
/**
* Recursively processes a memory file and all its @include references
* Returns an array of MemoryFileInfo objects with includes first, then main file
*/
export async function processMemoryFile(
filePath: string,
type: MemoryType,
processedPaths: Set<string>,
includeExternal: boolean,
depth: number = 0,
parent?: string,
): Promise<MemoryFileInfo[]> {
// Skip if already processed or max depth exceeded.
// Normalize paths for comparison to handle Windows drive letter casing
// differences (e.g., C:\Users vs c:\Users).
const normalizedPath = normalizePathForComparison(filePath)
if (processedPaths.has(normalizedPath) || depth >= MAX_INCLUDE_DEPTH) {
return []
}
// Skip if path is excluded by claudeMdExcludes setting
if (isClaudeMdExcluded(filePath, type)) {
return []
}
// Resolve symlink path early for @import resolution
const { resolvedPath, isSymlink } = safeResolvePath(
getFsImplementation(),
filePath,
)
processedPaths.add(normalizedPath)
if (isSymlink) {
processedPaths.add(normalizePathForComparison(resolvedPath))
}
const { info: memoryFile, includePaths: resolvedIncludePaths } =
await safelyReadMemoryFileAsync(filePath, type, resolvedPath)
if (!memoryFile || !memoryFile.content.trim()) {
return []
}
// Add parent information
if (parent) {
memoryFile.parent = parent
}
const result: MemoryFileInfo[] = []
// Add the main file first (parent before children)
result.push(memoryFile)
for (const resolvedIncludePath of resolvedIncludePaths) {
const isExternal = !pathInOriginalCwd(resolvedIncludePath)
if (isExternal && !includeExternal) {
continue
}
// Recursively process included files with this file as parent
const includedFiles = await processMemoryFile(
resolvedIncludePath,
type,
processedPaths,
includeExternal,
depth + 1,
filePath, // Pass current file as parent
)
result.push(...includedFiles)
}
return result
}
/**
* Processes all .md files in the .claude/rules/ directory and its subdirectories
* @param rulesDir The path to the rules directory
* @param type Type of memory file (User, Project, Local)
* @param processedPaths Set of already processed file paths
* @param includeExternal Whether to include external files
* @param conditionalRule If true, only include files with frontmatter paths; if false, only include files without frontmatter paths
* @param visitedDirs Set of already visited directory real paths (for cycle detection)
* @returns Array of MemoryFileInfo objects
*/
export async function processMdRules({
rulesDir,
type,
processedPaths,
includeExternal,
conditionalRule,
visitedDirs = new Set(),
}: {
rulesDir: string
type: MemoryType
processedPaths: Set<string>
includeExternal: boolean
conditionalRule: boolean
visitedDirs?: Set<string>
}): Promise<MemoryFileInfo[]> {
if (visitedDirs.has(rulesDir)) {
return []
}
try {
const fs = getFsImplementation()
const { resolvedPath: resolvedRulesDir, isSymlink } = safeResolvePath(
fs,
rulesDir,
)
visitedDirs.add(rulesDir)
if (isSymlink) {
visitedDirs.add(resolvedRulesDir)
}
const result: MemoryFileInfo[] = []
let entries: import('fs').Dirent[]
try {
entries = await fs.readdir(resolvedRulesDir)
} catch (e: unknown) {
const code = getErrnoCode(e)
if (code === 'ENOENT' || code === 'EACCES' || code === 'ENOTDIR') {
return []
}
throw e
}
for (const entry of entries) {
const entryPath = join(rulesDir, entry.name)
const { resolvedPath: resolvedEntryPath, isSymlink } = safeResolvePath(
fs,
entryPath,
)
// Use Dirent methods for non-symlinks to avoid extra stat calls.
// For symlinks, we need stat to determine what the target is.
const stats = isSymlink ? await fs.stat(resolvedEntryPath) : null
const isDirectory = stats ? stats.isDirectory() : entry.isDirectory()
const isFile = stats ? stats.isFile() : entry.isFile()
if (isDirectory) {
result.push(
...(await processMdRules({
rulesDir: resolvedEntryPath,
type,
processedPaths,
includeExternal,
conditionalRule,
visitedDirs,
})),
)
} else if (isFile && entry.name.endsWith('.md')) {
const files = await processMemoryFile(
resolvedEntryPath,
type,
processedPaths,
includeExternal,
)
result.push(
...files.filter(f => (conditionalRule ? f.globs : !f.globs)),
)
}
}
return result
} catch (error) {
if (error instanceof Error && error.message.includes('EACCES')) {
logEvent('tengu_claude_rules_md_permission_error', {
is_access_error: 1,
has_home_dir: rulesDir.includes(getClaudeConfigHomeDir()) ? 1 : 0,
})
}
return []
}
}
export const getMemoryFiles = memoize(
async (forceIncludeExternal: boolean = false): Promise<MemoryFileInfo[]> => {
const startTime = Date.now()
logForDiagnosticsNoPII('info', 'memory_files_started')
const result: MemoryFileInfo[] = []
const processedPaths = new Set<string>()
const config = getCurrentProjectConfig()
const includeExternal =
forceIncludeExternal ||
config.hasClaudeMdExternalIncludesApproved ||
false
// Process Managed file first (always loaded - policy settings)
const managedClaudeMd = getMemoryPath('Managed')
result.push(
...(await processMemoryFile(
managedClaudeMd,
'Managed',
processedPaths,
includeExternal,
)),
)
// Process Managed .claude/rules/*.md files
const managedClaudeRulesDir = getManagedClaudeRulesDir()
result.push(
...(await processMdRules({
rulesDir: managedClaudeRulesDir,
type: 'Managed',
processedPaths,
includeExternal,
conditionalRule: false,
})),
)
// Process User file (only if userSettings is enabled)
if (isSettingSourceEnabled('userSettings')) {
const userClaudeMd = getMemoryPath('User')
result.push(
...(await processMemoryFile(
userClaudeMd,
'User',
processedPaths,
true, // User memory can always include external files
)),
)
// Process User ~/.claude/rules/*.md files
const userClaudeRulesDir = getUserClaudeRulesDir()
result.push(
...(await processMdRules({
rulesDir: userClaudeRulesDir,
type: 'User',
processedPaths,
includeExternal: true,
conditionalRule: false,
})),
)
}
// Then process Project and Local files
const dirs: string[] = []
const originalCwd = getOriginalCwd()
let currentDir = originalCwd
while (currentDir !== parse(currentDir).root) {
dirs.push(currentDir)
currentDir = dirname(currentDir)
}
// When running from a git worktree nested inside its main repo (e.g.,
// .claude/worktrees/<name>/ from `claude -w`), the upward walk passes
// through both the worktree root and the main repo root. Both contain
// checked-in files like CLAUDE.md and .claude/rules/*.md, so the same
// content gets loaded twice. Skip Project-type (checked-in) files from
// directories above the worktree but within the main repo — the worktree
// already has its own checkout. CLAUDE.local.md is gitignored so it only
// exists in the main repo and is still loaded.
// See: https://github.com/anthropics/claude-code/issues/29599
const gitRoot = findGitRoot(originalCwd)
const canonicalRoot = findCanonicalGitRoot(originalCwd)
const isNestedWorktree =
gitRoot !== null &&
canonicalRoot !== null &&
normalizePathForComparison(gitRoot) !==
normalizePathForComparison(canonicalRoot) &&
pathInWorkingPath(gitRoot, canonicalRoot)
// Process from root downward to CWD
for (const dir of dirs.reverse()) {
// In a nested worktree, skip checked-in files from the main repo's
// working tree (dirs inside canonicalRoot but outside the worktree).
const skipProject =
isNestedWorktree &&
pathInWorkingPath(dir, canonicalRoot) &&
!pathInWorkingPath(dir, gitRoot)
// Try reading CLAUDE.md (Project) - only if projectSettings is enabled
if (isSettingSourceEnabled('projectSettings') && !skipProject) {
const projectPath = join(dir, 'CLAUDE.md')
result.push(
...(await processMemoryFile(
projectPath,
'Project',
processedPaths,
includeExternal,
)),
)
// Try reading .claude/CLAUDE.md (Project)
const dotClaudePath = join(dir, '.claude', 'CLAUDE.md')
result.push(
...(await processMemoryFile(
dotClaudePath,
'Project',
processedPaths,
includeExternal,
)),
)
// Try reading .claude/rules/*.md files (Project)
const rulesDir = join(dir, '.claude', 'rules')
result.push(
...(await processMdRules({
rulesDir,
type: 'Project',
processedPaths,
includeExternal,
conditionalRule: false,
})),
)
}
// Try reading CLAUDE.local.md (Local) - only if localSettings is enabled
if (isSettingSourceEnabled('localSettings')) {
const localPath = join(dir, 'CLAUDE.local.md')
result.push(
...(await processMemoryFile(
localPath,
'Local',
processedPaths,
includeExternal,
)),
)
}
}
// Process CLAUDE.md from additional directories (--add-dir) if env var is enabled
// This is controlled by CLAUDE_CODE_ADDITIONAL_DIRECTORIES_CLAUDE_MD and defaults to off
// Note: we don't check isSettingSourceEnabled('projectSettings') here because --add-dir
// is an explicit user action and the SDK defaults settingSources to [] when not specified
if (isEnvTruthy(process.env.CLAUDE_CODE_ADDITIONAL_DIRECTORIES_CLAUDE_MD)) {
const additionalDirs = getAdditionalDirectoriesForClaudeMd()
for (const dir of additionalDirs) {
// Try reading CLAUDE.md from the additional directory
const projectPath = join(dir, 'CLAUDE.md')
result.push(
...(await processMemoryFile(
projectPath,
'Project',
processedPaths,
includeExternal,
)),
)
// Try reading .claude/CLAUDE.md from the additional directory
const dotClaudePath = join(dir, '.claude', 'CLAUDE.md')
result.push(
...(await processMemoryFile(
dotClaudePath,
'Project',
processedPaths,
includeExternal,
)),
)
// Try reading .claude/rules/*.md files from the additional directory
const rulesDir = join(dir, '.claude', 'rules')
result.push(
...(await processMdRules({
rulesDir,
type: 'Project',
processedPaths,
includeExternal,
conditionalRule: false,
})),
)
}
}
// Memdir entrypoint (memory.md) - only if feature is on and file exists
if (isAutoMemoryEnabled()) {
const { info: memdirEntry } = await safelyReadMemoryFileAsync(
getAutoMemEntrypoint(),
'AutoMem',
)
if (memdirEntry) {
const normalizedPath = normalizePathForComparison(memdirEntry.path)
if (!processedPaths.has(normalizedPath)) {
processedPaths.add(normalizedPath)
result.push(memdirEntry)
}
}
}
// Team memory entrypoint - only if feature is on and file exists
if (feature('TEAMMEM') && teamMemPaths!.isTeamMemoryEnabled()) {
const { info: teamMemEntry } = await safelyReadMemoryFileAsync(
teamMemPaths!.getTeamMemEntrypoint(),
'TeamMem',
)
if (teamMemEntry) {