opencrabs 0.5.1

The autonomous, self-improving AI agent. Single Rust binary. Every channel. Recommended: the 40MB prebuilt binary for macOS, Linux and Windows: https://github.com/adolfousier/opencrabs/releases
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
# Spanish phantom-detection configuration
# Phrases are lower-case. Matched against lowercased response text.

intent_phrases = [
    "voy a cargar",
    "déjame cargar",
    "cargando el ",
    "cargando la ",
    "déjame revisar",
    "déjame verificar",
    "déjame inspeccionar",
    "déjame comprobar",
    "déjame actualizar",
    "déjame modificar",
    "déjame corregir",
    "déjame arreglar",
    "déjame crear",
    "déjame añadir",
    "déjame agregar",
    "déjame eliminar",
    "déjame borrar",
    "déjame ejecutar",
    "déjame leer",
    "déjame buscar",
    "déjame encontrar",
    "voy a revisar",
    "voy a verificar",
    "voy a actualizar",
    "voy a modificar",
    "voy a corregir",
    "voy a arreglar",
    "voy a crear",
    "voy a usar",
    "voy a utilizar",
    "voy a escribir",
    "voy a ejecutar",
    "voy a añadir",
    "voy a agregar",
    "voy a eliminar",
    "voy a borrar",
    "voy a ejecutar",
    "voy a leer",
    "voy a buscar",
    "ahora voy a",
    "ahora reviso",
    "ahora actualizo",
    "ahora corrijo",
    "ahora arreglo",
    "ahora creo",
    "ahora añado",
    "ahora elimino",
    "ahora ejecuto",
    "ahora leo",
    "ahora busco",
    "vamos a revisar",
    "vamos a actualizar",
    "vamos a corregir",
    "vamos a arreglar",
    "vamos a crear",
    "vamos a añadir",
    "vamos a eliminar",
    "vamos a ejecutar",
    "procedo a",
    "procederé a",
    "ahora procedo a",

    # Deferment stalls — "necesito X" / "tengo que X" / "debo X"
    # 2026-05-28: parity with the English "i need to X" entries; this is
    # the same pre-action narration shape in Spanish.
    "necesito leer",
    "necesito revisar",
    "necesito verificar",
    "necesito comprobar",
    "necesito inspeccionar",
    "necesito examinar",
    "necesito buscar",
    "necesito encontrar",
    "necesito entender",
    "necesito ver",
    "necesito investigar",
    "necesito actualizar",
    "necesito corregir",
    "necesito arreglar",
    "necesito modificar",
    "necesito crear",
    "necesito escribir",
    "necesito añadir",
    "necesito agregar",
    "necesito eliminar",
    "tengo que leer",
    "tengo que revisar",
    "tengo que verificar",
    "tengo que actualizar",
    "tengo que corregir",
    "tengo que arreglar",
    "tengo que investigar",
    "debo leer",
    "debo revisar",
    "debo verificar",
    "debo actualizar",
    "debo corregir",
    "debería leer",
    "debería revisar",
    "debería verificar",
    "debería investigar",
    # Cleanup / destructive action verbs — 2026-06-04: mirror of the
    # EN additions (limpiar / eliminar / borrar / quitar /
    # refactorizar / reorganizar / ordenar / purgar). Spanish
    # conjugates first-person in the verb itself, so the equivalents
    # of "let me X" / "I'll X" / pronounless "need to X" collapse
    # into the standard set of prefixes the bot already uses.
    "déjame limpiar",
    "déjame eliminar",
    "déjame borrar",
    "déjame quitar",
    "déjame refactorizar",
    "déjame reorganizar",
    "déjame ordenar",
    "déjame purgar",
    "voy a limpiar",
    "voy a eliminar",
    "voy a borrar",
    "voy a quitar",
    "voy a refactorizar",
    "voy a reorganizar",
    "voy a ordenar",
    "voy a purgar",
    "vamos a limpiar",
    "vamos a eliminar",
    "vamos a borrar",
    "vamos a quitar",
    "vamos a refactorizar",
    "vamos a reorganizar",
    "vamos a ordenar",
    "vamos a purgar",
    "necesito limpiar",
    "necesito eliminar",
    "necesito borrar",
    "necesito quitar",
    "necesito refactorizar",
    "necesito reorganizar",
    "necesito ordenar",
    "necesito purgar",
    "tengo que limpiar",
    "tengo que eliminar",
    "tengo que borrar",
    "tengo que quitar",
    "tengo que refactorizar",
    "tengo que reorganizar",
    "tengo que ordenar",
    "tengo que purgar",
    "debo limpiar",
    "debo eliminar",
    "debo borrar",
    "debo refactorizar",
    "debería limpiar",
    "debería eliminar",
    "debería borrar",
    "debería refactorizar",
    "debería reorganizar",
]

action_verbs = [
    # Local playback / launching an artifact (see en.toml).
    "reproduje",
    "abrí el archivo",
    "lancé",
    "previsualicé",
    "actualicé",
    "modifiqué",
    "corregí",
    "arreglé",
    "creé",
    "añadí",
    "agregué",
    "eliminé",
    "borré",
    "ejecuté",
    "leí",
    "busqué",
    "encontré",
    "revisé",
    "verifiqué",
    "inspeccioné",
    "comprobé",
    "actualizado",
    "modificado",
    "corregido",
    "arreglado",
    "creado",
    "añadido",
    "agregado",
    "eliminado",
    "borrado",
    "ejecutado",
    # Issue-tracker and review actions (see en.toml for the case that
    # exposed the gap).
    "cerré",
    "cerrado",
    "reabrí",
    "reabierto",
    "comenté",
    "comentado",
    "abrí",
    "abierto",
    "etiqueté",
    "asigné",
    # Self-update and lifecycle. A fabricated "Evolved from X to Y" with zero
    # tool calls told the user their binary had replaced itself when it had
    # not, leaving them to report bugs against a version they were not running.
    "evolucioné",
    "evolucionado",
    "actualicé la versión",
    "reinicié",
    "reiniciado",
    "reconstruí",
    # Claimed INVESTIGATION, not side effects. A zero-tool turn opening
    # "Verified just now, real output:" produced a fabricated file dump with
    # invented line numbers. Fabricated evidence is worse than a fabricated
    # action: a fake push dies to one git log, a fake dump gets built on.
    # Deliberately excludes read/checked/found/ran — all legitimate in
    # ordinary conversation about context the model already has.
    "verificado",
    "audité",
    "auditado",
]

completion_claims = [
    "archivo actualizado",
    "archivo modificado",
    "archivo corregido",
    "archivo creado",
    "código actualizado",
    "código corregido",
    "error corregido",
    "cambios aplicados",
    "todo actualizado",
    "todo corregido",
    "listo",
    "hecho",
    "completado",
    "guardado",
    "el trabajo está entregado",
    "trabajo entregado y registrado",
]

line_start_re = "(?im)^[\\s\\-*•>]*(?:(?:vale|bien|ok|okey|sí|claro)\\s*,?\\s+)?(?:ahora\\s+)?(?:déjame|voy a|vamos a|necesito|tengo que|debo|debería|debería de|debo de)\\s+\\w"

# Anuncio de trabajo en curso que cierra el turno sin llamar a ninguna
# herramienta — "Ejecutando las pruebas ahora.", "Verificando…". Requiere
# gerundio de acción + marcador de inmediatez (ahora / … / : ).
work_announcement_re = "(?i)^[\\s\\-*•>]*(?:(?:(?:y\\s+)?luego|(?:y\\s+)?despu[ée]s|(?:y\\s+)?entonces|a\\s+continuaci[óo]n|y)[,]?\\s+)?(?:(?:vale|bien|ok|okey|sí|claro)[,.!:]?\\s+)?(?:estoy\\s+|ahora\\s+)?(?:ejecutando|corriendo|verificando|revisando|comprobando|construyendo|compilando|probando|buscando|leyendo|escribiendo|creando|generando|actualizando|modificando|instalando|aplicando|desplegando|descargando|subiendo|eliminando|borrando|arreglando|corrigiendo|analizando|investigando)\\b[^.!?:\\n]{0,60}?(?:\\s+ahora(?:[.!…,:]+|\\s*(?:\\n|$))|\\.{3}|…|:\\s*(?:\\n|$))"
# Gerund-led plan announcement carrying NO imminence marker — "Setting up the
# plan, then mapping every call site before touching anything." Two reasons
# the patterns above cannot see it (#1261): the verb is a PREPARATION verb
# (mapping, planning, drafting, listing) that work_announcement_re never
# enumerated, and the clause is sequenced with then / before instead of ending
# in now / … / :, which that pattern requires and gerund_re expects at the
# front. Scanned by the zero-tool gate only: once a tool has run, the same
# shape can be a legitimate recap ("Updating the config fixed it, then the
# build passed").
plan_announcement_re = "(?im)^[\\s\\-*•>]*(?:(?:ok|vale|bien|claro)[,.!:]?\\s+)?(?:voy\\s+a\\s+|estoy\\s+|primero\\s+|ahora\\s+)?(?:configurando|configurar|mapeando|mapear|planificando|planificar|preparando|preparar|revisando|revisar|listando|listar|documentando|documentar|creando|crear|actualizando|actualizar|verificando|verificar|leyendo|leer|ejecutando|ejecutar|analizando|analizar|investigando|investigar)\\b[^.!?\\n]{0,120}?\\b(?:luego|después|despues|entonces|antes\\s+de|a\\s+continuación|a\\s+continuacion)\\b"


gerund_re = "(?im)(?:^|[.!?]\\s+)\\s*(?:ahora|ya)\\s+(?:actualizando|modificando|corrigiendo|arreglando|creando|añadiendo|agregando|eliminando|borrando|ejecutando|leyendo|buscando)\\b"

trailing_colon_re = "(?im)(?:^|\\n)\\s*(?:déjame|voy a|ahora voy a|vamos a)\\s+\\w[^:\\n]{0,80}:\\s*$"

now_imperative_re = "(?m)^[\\s\\-*]*(?:ahora\\s+)?(?:déjame|voy a|vamos a)\\s+\\w"

numbered_steps_re = "(?m)^\\s*\\d+\\.\\s+(?:actualizar|modificar|corregir|arreglar|crear|añadir|agregar|eliminar|borrar|ejecutar|leer|buscar)"

past_tense_standalone_re = "(?m)^[\\s\\-*]*(?:actualicé|modifiqué|corregí|arreglé|creé|añadí|agregué|eliminé|borré|ejecuté|revisé|verifiqué|actualizado|modificado|corregido|arreglado|creado)[.!]"

path_re = "(?:^|[\\s`(])(?:\\./)?[a-zA-Z_][\\w\\-]*/[\\w\\-/]*\\.\\w{1,6}(?:[\\s`),:;]|$)"

ext_re = "(?:^|[\\s`(])[\\w\\-]+\\.(?:rs|py|ts|tsx|js|jsx|go|sh|toml|yaml|yml|json|md)(?:[\\s`),:;]|$)"

backtick_code_re = "`[a-zA-Z_]\\w+`"

# Alucinación de generación de imágenes (#747)
media_delivery_phrases = [
    "aquí está",
    "aquí lo tienes",
    "aquí tienes",
    "ya está",
    "listo, aquí",
    "lo generé",
    "he generado",
    "generé la imagen",
    "creé la imagen",
    "edité la imagen",
    "la imagen editada",
    "la imagen generada",
    "generado esta vez",
    "generada esta vez",
    "envié la imagen",
]

media_context_words = [
    "imagen",
    "foto",
    "fotografía",
    "render",
    "fondo",
    "brillo",
    "exposición",
    "contraste",
    "logo",
    "logotipo",
]

# Framings that present a named command as ALREADY RUN. The command is
# language-neutral; the claim around it is not. See #789.
executed_framings = [
    "ejecuté",
    "acabo de ejecutar",
    "corrí",
    "verifiqué con",
    "comprobé con",
    "la salida",
    "salida arriba",
    "el resultado",
    "confirmé con",
]

# Claims that a FILE or DOCUMENT was sent (#825). Paired with
# `file_context_words`: a claim of a delivered file in a turn that never
# invoked a document-sending tool is fabricating. Document sending is a
# newer capability than image generation, so the media checks (#747) did
# not cover it.
file_delivery_phrases = [
    "archivo enviado",
    "envié el archivo",
    "te envié el archivo",
    "adjunté",
    "he adjuntado",
    "adjunto el archivo",
    "subí el archivo",
    "he subido",
    "documento enviado",
    "envié el documento",
    "envié el informe",
    "informe entregado",
    "el archivo está arriba",
    "ver el adjunto",
    "recién salido del horno",
    "aquí está el vídeo",
    "aquí está el audio",
    "está ahí arriba",
    "ahí arriba",
    "dale play",
    "acabo de enviarlo",
    "ya lo envié",
]

# Words marking the claimed deliverable as a file rather than a message.
file_context_words = [
    "archivo",
    "documento",
    "informe",
    ".md",
    "adjunto",
    "pdf",
    "csv",
    ".mp4",
    ".mp3",
    ".wav",
    ".png",
    ".jpg",
    "vídeo",
    "audio",
    "imagen",
    "mp4",
    "mp3",
    "wav",
    "mov",
    "png",
    "jpg",
]

# Generic first-person intent: the verb is captured, not enumerated, so a
# verb missing from intent_phrases (e.g. "let me execute") is still caught.
generic_intent_re = "(?i)\\b(?:voy\\s+a|d[eé]jame|perm[ií]teme|vamos\\s+a)\\s+([a-záéíóúñ]{2,})"

# Verbs that follow that construction without promising any work.
intent_verb_exclusions = [
    "saber",
    "explicar",
    "aclarar",
    "resumir",
    "decir",
    "pensar",
    "mostrar",
    "enseñar",
    "compartir",
    "describir",
    "mencionar",
    "confirmar",
]


# Context that turns a BARE tool name into a claim of tool usage (#1262).
# Multi-word names (`load_brain_file`) are unmistakable on their own, but
# `plan`, `bash`, `grep` and `ls` are ordinary words, so naming one proves
# nothing by itself. It proves something when the name is the OBJECT of an
# action the model attributes to itself: "Setting up the plan" in a turn that
# called no tool is a tool use that never happened. A marker must sit
# immediately before the name, optionally across one determiner, so an
# instruction addressed to the user ("run it in bash") stays untouched.
tool_use_markers = [
    "usando",
    "usar",
    "uso",
    "usé",
    "utilizando",
    "utilizar",
    "ejecutando",
    "ejecutar",
    "ejecuté",
    "corriendo",
    "correr",
    "llamando",
    "llamar",
    "llamé",
    "invocando",
    "invocar",
    "invoqué",
    "creando",
    "crear",
    "creé",
    "configurando",
    "configurar",
    "configuré",
    "preparando",
    "preparar",
    "actualizando",
    "actualizar",
    "actualicé",
    "leyendo",
    "leer",
    "leí",
    "escribiendo",
    "escribir",
    "verificando",
    "verificar",
    "verifiqué",
    "abriendo",
    "abrir",
    "abrí",
    "iniciando",
    "iniciar",
    "vía",
    "via",
]

# Nouns that mark the preceding word as a tool by name ("the plan tool").
tool_nouns = [
    "herramienta",
    "herramientas",
]

# Determiners allowed between a marker and the tool name, so "setting up THE
# plan" reads the same as "setting up plan". At most one, and only these:
# anything else between the two words means the name is not the marker's
# object. Empty for languages without articles.
tool_name_determiners = [
    "el",
    "la",
    "los",
    "las",
    "un",
    "una",
    "mi",
    "nuestro",
    "nuestra",
    "este",
    "esta",
    "del",
]