{{ message }}
Repository navigation
Expand file tree
/
Copy pathcheck-actorless-executor-operations.ts
More file actions
275 lines (245 loc) · 11.4 KB
/
Copy pathcheck-actorless-executor-operations.ts
File metadata and controls
275 lines (245 loc) · 11.4 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
#!/usr/bin/env bun
/**
* Fails when an operation an actorless run can reach demands a human subject.
*
* A workflow executes under a `Principal`, and several triggers have no person
* behind them: a schedule, the public API, a webhook carrying no external subject.
* Those runs still hold real authority — `workspace-authorization.ts` admits an
* actorless executor delegation whose current workflow is a deployment — so they
* are authorized callers with no `subjectUserId`. Any use case they can reach that
* calls `requirePrincipalSubjectUserId` therefore throws, and because
* `PrincipalSubjectUserRequiredError` is not an `OrchestrationError` it surfaces as
* an opaque 500 rather than anything a workflow author can act on.
*
* That is not hypothetical: the Logs detail tools broke for every scheduled run
* this way, and the failure reached production because nothing connected "this
* operation admits `delegatedServices: ['executor']`" to "this use case requires a
* person". This audit connects them.
*
* A subject is genuinely required often enough that the rule is an annotation, not
* a ban. Write `// actorless-unsupported: <reason>` above the call to declare that
* the operation has no meaning without a person — the annotation turns a silent
* 500 into a documented gap that a reviewer can weigh.
*/
import { readdirSync, readFileSync, statSync } from 'node:fs'
import { dirname, join, relative, resolve } from 'node:path'
import { fileURLToPath } from 'node:url'
const SCRIPT_DIR = dirname(fileURLToPath(import.meta.url))
const ROOT = resolve(SCRIPT_DIR, '..')
const SCAN_ROOTS = ['apps/sim/lib', 'apps/sim/app']
const REQUIRE_CALL = 'requirePrincipalSubjectUserId('
const ANNOTATION = 'actorless-unsupported:'
const MAX_ANNOTATION_LOOKBACK = 3
/**
* `apps/sim/lib/internal/**` is the in-process tool surface: every handler under it
* mints an executor delegation, so its modules are executor-reachable whether or
* not they bind a named operation.
*/
const EXECUTOR_SURFACE_PREFIX = 'apps/sim/lib/internal/'
export interface ActorlessFinding {
file: string
line: number
reason: 'unannotated' | 'empty-reason'
operations: string[]
}
/** How the file was shown to be reachable by an actorless run. */
type Reachability = 'internal-surface' | 'declared-operation' | 'unproven'
/** Text of the balanced `(...)` or `{...}` group that starts at `openIndex`. */
function balancedGroup(source: string, openIndex: number): string {
const open = source[openIndex]
const close = open === '(' ? ')' : '}'
let depth = 0
for (let index = openIndex; index < source.length; index++) {
const char = source[index]
if (char === open) depth++
else if (char === close) {
depth--
if (depth === 0) return source.slice(openIndex, index + 1)
}
}
return source.slice(openIndex)
}
function admitsExecutor(text: string): boolean {
const match = /delegatedServices\s*:\s*\[([^\]]*)\]/.exec(text)
return match ? /['"]executor['"]/.test(match[1]) : false
}
/**
* Maps every `<namespace>.<key>` declared in an operations module to whether its
* policy admits an executor delegation, resolving same-file policy constants that
* are spread into the definition (e.g. `...LOG_READER_PRINCIPAL_POLICY`).
*/
export function parseOperationPolicies(source: string): Map<string, boolean> {
const policies = new Map<string, boolean>()
const spreadable = new Map<string, boolean>()
const constPattern = /(?:^|\n)\s*(?:export\s+)?const\s+([A-Za-z0-9_$]+)\s*=\s*\{/g
for (let match = constPattern.exec(source); match; match = constPattern.exec(source)) {
const braceIndex = source.indexOf('{', match.index + match[0].length - 1)
spreadable.set(match[1], admitsExecutor(balancedGroup(source, braceIndex)))
}
/** Whether a `defineWorkspaceOperation({...})` call admits an executor delegation. */
const definitionAdmitsExecutor = (definition: string): boolean => {
if (admitsExecutor(definition)) return true
return [...definition.matchAll(/\.\.\.([A-Za-z0-9_$]+)/g)].some(
(spread) => spreadable.get(spread[1]) === true
)
}
// Several domains declare their operations through same-file factories
// (`toolReadOperation('tables.rows.query')`) rather than inline, so the policy has
// to be resolved through the factory or those operations read as executor-free.
const factories = new Map<string, boolean>()
const factoryPattern = /(?:^|\n)\s*(?:export\s+)?function\s+([A-Za-z0-9_$]+)\s*[<(]/g
for (let match = factoryPattern.exec(source); match; match = factoryPattern.exec(source)) {
const bodyIndex = source.indexOf('{', match.index + match[0].length - 1)
if (bodyIndex === -1) continue
const body = balancedGroup(source, bodyIndex)
const defineIndex = body.indexOf('defineWorkspaceOperation')
if (defineIndex === -1) continue
const parenIndex = body.indexOf('(', defineIndex)
factories.set(match[1], definitionAdmitsExecutor(balancedGroup(body, parenIndex)))
}
const namespacePattern = /(?:^|\n)\s*export\s+const\s+([A-Za-z0-9_$]+)\s*=\s*\{/g
for (let match = namespacePattern.exec(source); match; match = namespacePattern.exec(source)) {
const namespace = match[1]
const braceIndex = source.indexOf('{', match.index + match[0].length - 1)
const body = balancedGroup(source, braceIndex)
const entryPattern = /([A-Za-z0-9_$]+)\s*:\s*([A-Za-z0-9_$]+)\s*\(/g
for (let entry = entryPattern.exec(body); entry; entry = entryPattern.exec(body)) {
const [, key, callee] = entry
if (callee === 'defineWorkspaceOperation') {
const parenIndex = body.indexOf('(', entry.index + entry[0].length - 1)
policies.set(
`${namespace}.${key}`,
definitionAdmitsExecutor(balancedGroup(body, parenIndex))
)
} else if (factories.has(callee)) {
policies.set(`${namespace}.${key}`, factories.get(callee) === true)
}
}
}
return policies
}
/** The declared operations a module references, whether to define or to bind them. */
export function referencedOperations(source: string, known: Set<string>): string[] {
const referenced = new Set<string>()
for (const match of source.matchAll(/([A-Za-z0-9_$]+)\.([A-Za-z0-9_$]+)/g)) {
const id = `${match[1]}.${match[2]}`
if (known.has(id)) referenced.add(id)
}
return [...referenced].sort()
}
/**
* Flags `requirePrincipalSubjectUserId` calls that are not declared actorless-unsupported.
* Mirrors the placement rule the boundary annotations use: the annotation must sit in one
* of the preceding comment lines, so extra context above it is fine.
*/
export function auditSubjectRequirements(source: string, operations: string[]): ActorlessFinding[] {
const findings: ActorlessFinding[] = []
const lines = source.split('\n')
for (const [index, line] of lines.entries()) {
if (!line.includes(REQUIRE_CALL)) continue
let annotation: string | undefined
for (let back = index - 1; back >= 0 && back >= index - MAX_ANNOTATION_LOOKBACK; back--) {
const candidate = lines[back].trim()
if (candidate === '') continue
if (!candidate.startsWith('//') && !candidate.startsWith('*')) break
const found = candidate.indexOf(ANNOTATION)
if (found !== -1) {
annotation = candidate.slice(found + ANNOTATION.length).trim()
break
}
}
if (annotation === undefined) {
findings.push({ file: '', line: index + 1, reason: 'unannotated', operations })
} else if (annotation === '') {
findings.push({ file: '', line: index + 1, reason: 'empty-reason', operations })
}
}
return findings
}
function walk(directory: string, into: string[]): string[] {
for (const entry of readdirSync(directory)) {
if (entry === 'node_modules' || entry === '.next') continue
const full = join(directory, entry)
if (statSync(full).isDirectory()) walk(full, into)
else if (full.endsWith('.ts') && !full.endsWith('.test.ts')) into.push(full)
}
return into
}
function main(): void {
const sourceFiles = SCAN_ROOTS.flatMap((root) => walk(join(ROOT, root), []))
const policies = new Map<string, boolean>()
for (const file of sourceFiles) {
if (!file.endsWith('/application/operations.ts')) continue
for (const [id, executor] of parseOperationPolicies(readFileSync(file, 'utf8'))) {
policies.set(id, executor)
}
}
const known = new Set(policies.keys())
const executorAdmitting = new Set([...policies].filter(([, yes]) => yes).map(([id]) => id))
// Domains with at least one executor-admitting operation. A shared use-case
// factory in such a domain names no operation of its own — it takes one as an
// argument — so it cannot be proven executor-free and fails closed here.
const executorDomains = new Set<string>()
for (const file of sourceFiles) {
if (!file.endsWith('/application/operations.ts')) continue
const parsed = parseOperationPolicies(readFileSync(file, 'utf8'))
if ([...parsed.values()].some(Boolean)) {
executorDomains.add(relative(ROOT, dirname(dirname(file))))
}
}
const findings: ActorlessFinding[] = []
const reachabilityByFinding = new Map<ActorlessFinding, Reachability>()
let auditedFiles = 0
for (const file of sourceFiles) {
const source = readFileSync(file, 'utf8')
if (!source.includes(REQUIRE_CALL)) continue
const relativePath = relative(ROOT, file)
const operations = referencedOperations(source, known)
const reachability: Reachability | undefined = relativePath.startsWith(EXECUTOR_SURFACE_PREFIX)
? 'internal-surface'
: operations.some((id) => policies.get(id) === true)
? 'declared-operation'
: operations.length === 0 &&
[...executorDomains].some((domain) => relativePath.startsWith(`${domain}/`))
? 'unproven'
: undefined
if (!reachability) continue
auditedFiles++
for (const audited of auditSubjectRequirements(source, operations)) {
const finding = { ...audited, file: relativePath }
findings.push(finding)
reachabilityByFinding.set(finding, reachability)
}
}
if (findings.length > 0) {
console.error(
'Operations an actorless run can reach must not silently require a human subject:'
)
for (const finding of findings) {
const reachability = reachabilityByFinding.get(finding)
const via =
reachability === 'internal-surface'
? 'in-process tool surface'
: reachability === 'unproven'
? 'shared use case in a domain with executor-admitting operations'
: `reachable via ${finding.operations.filter((id) => executorAdmitting.has(id)).join(', ')}`
const problem =
finding.reason === 'empty-reason'
? `${ANNOTATION} needs a reason`
: `unannotated requirePrincipalSubjectUserId (${via})`
console.error(` ${finding.file}:${finding.line} ${problem}`)
}
console.error(
`\nA scheduled, public-API, or subject-less webhook run reaches these with no user, and` +
`\n\`requirePrincipalSubjectUserId\` throws a 500 there rather than anything actionable.` +
`\nEither resolve the user optionally (\`resolvePrincipalSubjectUserId\`) when it is only` +
`\nattribution, or declare the gap with \`// ${ANNOTATION} <reason>\` above the call.`
)
process.exit(1)
}
console.log(
`✓ no undeclared human-subject requirements on actorless-reachable operations ` +
`(${executorAdmitting.size} executor-admitting operations, ${auditedFiles} files audited)`
)
}
if (import.meta.main) main()
You can’t perform that action at this time.