1use std::ffi::{OsStr, OsString};
11use std::fmt;
12
13const REGISTERED_NAMES: &str = include_str!("environment_registry_names.txt");
14const UNKNOWN_CODE: &str = "HARN-ENV-001";
15const INVALID_VALUE_CODE: &str = "HARN-ENV-002";
16
17#[derive(Clone, Copy, Debug, PartialEq, Eq)]
19pub enum EnvironmentConsumer {
20 Runtime,
21 Cli,
22 BuildTooling,
23 TestHarness,
24 EmbedderExtension,
25}
26
27#[derive(Clone, Copy, Debug, PartialEq, Eq)]
32pub enum EnvironmentValueShape {
33 OwnerValidated,
34 Boolean,
35 UnsignedInteger,
36 NonNegativeNumber,
37 UnitInterval,
38 Enumerated(&'static [&'static str]),
39}
40
41#[derive(Clone, Copy, Debug, PartialEq, Eq)]
43pub enum EnvironmentSensitivity {
44 Public,
45 Credential,
46}
47
48#[derive(Clone, Debug, PartialEq, Eq)]
49pub struct EnvironmentVariableSpec {
50 pub name: String,
51 pub consumer: EnvironmentConsumer,
52 pub value_shape: EnvironmentValueShape,
53 pub sensitivity: EnvironmentSensitivity,
54}
55
56#[derive(Clone, Debug, PartialEq, Eq)]
57pub enum EnvironmentDiagnosticKind {
58 UnknownName { suggestion: Option<String> },
59 InvalidValue { expected: EnvironmentValueShape },
60}
61
62#[derive(Clone, Debug, PartialEq, Eq)]
65pub struct EnvironmentDiagnostic {
66 pub code: &'static str,
67 pub key: String,
68 pub kind: EnvironmentDiagnosticKind,
69}
70
71impl fmt::Display for EnvironmentDiagnostic {
72 fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
73 match &self.kind {
74 EnvironmentDiagnosticKind::UnknownName { suggestion } => {
75 write!(
76 formatter,
77 "{}: unknown Harn environment variable `{}`.",
78 self.code, self.key
79 )?;
80 if let Some(suggestion) = suggestion {
81 write!(formatter, " Did you mean `{suggestion}`?")?;
82 }
83 formatter.write_str(" Use `HARN_EXT_<NAME>` for settings owned by a calling tool.")
84 }
85 EnvironmentDiagnosticKind::InvalidValue { expected } => write!(
86 formatter,
87 "{}: environment variable `{}` must be {}",
88 self.code,
89 self.key,
90 expected.description()
91 ),
92 }
93 }
94}
95
96#[derive(Clone, Debug, PartialEq, Eq)]
97pub struct EnvironmentValidationError {
98 diagnostics: Vec<EnvironmentDiagnostic>,
99}
100
101impl EnvironmentValidationError {
102 pub fn diagnostics(&self) -> &[EnvironmentDiagnostic] {
103 &self.diagnostics
104 }
105}
106
107impl fmt::Display for EnvironmentValidationError {
108 fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
109 for (index, diagnostic) in self.diagnostics.iter().enumerate() {
110 if index > 0 {
111 formatter.write_str("\n")?;
112 }
113 diagnostic.fmt(formatter)?;
114 }
115 Ok(())
116 }
117}
118
119impl std::error::Error for EnvironmentValidationError {}
120
121impl EnvironmentValueShape {
122 fn description(self) -> String {
123 match self {
124 Self::OwnerValidated => "valid for its owning subsystem".to_string(),
125 Self::Boolean => {
126 "a boolean (`true`, `false`, `yes`, `no`, `on`, `off`, `1`, or `0`)".to_string()
127 }
128 Self::UnsignedInteger => "an unsigned integer".to_string(),
129 Self::NonNegativeNumber => "a non-negative number".to_string(),
130 Self::UnitInterval => "a number between 0 and 1".to_string(),
131 Self::Enumerated(values) => format!(
132 "one of {}",
133 values
134 .iter()
135 .map(|value| format!("`{value}`"))
136 .collect::<Vec<_>>()
137 .join(", ")
138 ),
139 }
140 }
141
142 fn accepts(self, value: &OsStr) -> bool {
143 let Some(value) = value.to_str() else {
144 return matches!(self, Self::OwnerValidated);
145 };
146 match self {
147 Self::OwnerValidated => true,
148 Self::Boolean => matches!(
149 value.trim().to_ascii_lowercase().as_str(),
150 "1" | "0" | "true" | "false" | "yes" | "no" | "on" | "off"
151 ),
152 Self::UnsignedInteger => value.trim().parse::<u64>().is_ok(),
153 Self::NonNegativeNumber => value
154 .trim()
155 .parse::<f64>()
156 .is_ok_and(|parsed| parsed.is_finite() && parsed >= 0.0),
157 Self::UnitInterval => value
158 .trim()
159 .parse::<f64>()
160 .is_ok_and(|parsed| parsed.is_finite() && (0.0..=1.0).contains(&parsed)),
161 Self::Enumerated(values) => {
162 let value = value.trim().to_ascii_lowercase();
163 values.contains(&value.as_str())
164 }
165 }
166 }
167}
168
169pub fn variable_spec(name: &str) -> Option<EnvironmentVariableSpec> {
171 if registered_names().binary_search(&name).is_ok() {
172 return Some(spec_for_registered_name(name));
173 }
174 if is_extension_name(name) {
175 return Some(EnvironmentVariableSpec {
176 name: name.to_string(),
177 consumer: EnvironmentConsumer::EmbedderExtension,
178 value_shape: EnvironmentValueShape::OwnerValidated,
179 sensitivity: sensitivity_for(name),
180 });
181 }
182 if is_structured_runtime_name(name) {
183 return Some(spec_for_registered_name(name));
184 }
185 None
186}
187
188pub fn validate_startup_environment() -> Result<(), EnvironmentValidationError> {
191 validate_environment(std::env::vars_os())
192}
193
194pub fn validate_environment<I, K, V>(vars: I) -> Result<(), EnvironmentValidationError>
197where
198 I: IntoIterator<Item = (K, V)>,
199 K: Into<OsString>,
200 V: Into<OsString>,
201{
202 let mut diagnostics = Vec::new();
203 for (key, value) in vars {
204 let key = key.into();
205 let Some(key) = key.to_str() else {
206 continue;
207 };
208 if !key.starts_with("HARN_") {
209 continue;
210 }
211 let Some(spec) = variable_spec(key) else {
212 diagnostics.push(EnvironmentDiagnostic {
213 code: UNKNOWN_CODE,
214 key: key.to_string(),
215 kind: EnvironmentDiagnosticKind::UnknownName {
216 suggestion: nearest_registered_name(key),
217 },
218 });
219 continue;
220 };
221 let value = value.into();
222 if !spec.value_shape.accepts(&value) {
223 diagnostics.push(EnvironmentDiagnostic {
224 code: INVALID_VALUE_CODE,
225 key: key.to_string(),
226 kind: EnvironmentDiagnosticKind::InvalidValue {
227 expected: spec.value_shape,
228 },
229 });
230 }
231 }
232 diagnostics.sort_by(|left, right| left.key.cmp(&right.key));
233 if diagnostics.is_empty() {
234 Ok(())
235 } else {
236 Err(EnvironmentValidationError { diagnostics })
237 }
238}
239
240fn registered_names() -> &'static [&'static str] {
241 static NAMES: std::sync::OnceLock<Vec<&'static str>> = std::sync::OnceLock::new();
242 NAMES
243 .get_or_init(|| {
244 REGISTERED_NAMES
245 .lines()
246 .filter(|name| !name.is_empty())
247 .collect()
248 })
249 .as_slice()
250}
251
252fn spec_for_registered_name(name: &str) -> EnvironmentVariableSpec {
253 EnvironmentVariableSpec {
254 name: name.to_string(),
255 consumer: consumer_for(name),
256 value_shape: value_shape_for(name),
257 sensitivity: sensitivity_for(name),
258 }
259}
260
261fn consumer_for(name: &str) -> EnvironmentConsumer {
262 if name.starts_with("HARN_TEST_") || name.contains("_TEST_") || name.starts_with("HARN_E2E_") {
263 EnvironmentConsumer::TestHarness
264 } else if [
265 "HARN_BUILD_",
266 "HARN_CARGO_",
267 "HARN_CHECK_",
268 "HARN_CI_",
269 "HARN_CODEGEN_",
270 "HARN_DEV_",
271 "HARN_RELEASE_",
272 ]
273 .iter()
274 .any(|prefix| name.starts_with(prefix))
275 {
276 EnvironmentConsumer::BuildTooling
277 } else if name.starts_with("HARN_CLI_")
278 || name.starts_with("HARN_DOCTOR_")
279 || name.starts_with("HARN_INIT_")
280 {
281 EnvironmentConsumer::Cli
282 } else {
283 EnvironmentConsumer::Runtime
284 }
285}
286
287fn value_shape_for(name: &str) -> EnvironmentValueShape {
288 match name {
289 "HARN_LLM_TIMEOUT"
290 | "HARN_LLM_IDLE_TIMEOUT"
291 | "HARN_LLM_FIRST_TOKEN_TIMEOUT"
292 | "HARN_MAX_CONCURRENCY"
293 | "HARN_RETENTION_DAYS"
294 | "HARN_TOKEN_BUDGET"
295 | "HARN_EVENT_LOG_QUEUE_DEPTH" => EnvironmentValueShape::UnsignedInteger,
296 "HARN_BUDGET_USD" => EnvironmentValueShape::NonNegativeNumber,
297 "HARN_OTEL_SAMPLE_RATIO" => EnvironmentValueShape::UnitInterval,
298 "HARN_BYTECODE_CACHE"
299 | "HARN_FLIGHT_RECORDER"
300 | harn_parser::HARN_LEGACY_AMBIENT_CAPABILITIES_ENV
301 | "HARN_LLM_STREAM"
302 | "HARN_REPLAY_ENABLED"
303 | "HARN_REQUIRE_SIGNED_SKILLS"
304 | "HARN_TRACE"
305 | "HARN_VERBOSE_CONFIG" => EnvironmentValueShape::Boolean,
306 crate::vm::subtask::PLACEMENT_ENV => {
307 EnvironmentValueShape::Enumerated(crate::vm::subtask::PLACEMENT_VALUES)
308 }
309 _ => EnvironmentValueShape::OwnerValidated,
310 }
311}
312
313fn sensitivity_for(name: &str) -> EnvironmentSensitivity {
314 if [
315 "TOKEN",
316 "SECRET",
317 "PASSWORD",
318 "API_KEY",
319 "OAUTH_KEY",
320 "HEADERS",
321 "PRIVATE_KEY",
322 ]
323 .iter()
324 .any(|fragment| name.contains(fragment))
325 {
326 EnvironmentSensitivity::Credential
327 } else {
328 EnvironmentSensitivity::Public
329 }
330}
331
332fn is_extension_name(name: &str) -> bool {
336 name.strip_prefix("HARN_EXT_")
337 .is_some_and(is_upper_identifier)
338}
339
340fn is_structured_runtime_name(name: &str) -> bool {
344 is_secret_name(name)
345 || is_rate_limit_name(name)
346 || is_model_role_name(name)
347 || is_agent_model_option_name(name)
348}
349
350fn is_secret_name(name: &str) -> bool {
351 name.strip_prefix("HARN_SECRET_")
352 .is_some_and(is_upper_identifier)
353}
354
355fn is_rate_limit_name(name: &str) -> bool {
356 let Some(suffix) = name.strip_prefix("HARN_RATE_LIMIT_") else {
357 return false;
358 };
359 let Some((provider, field)) = suffix.rsplit_once('_') else {
360 return false;
361 };
362 is_upper_identifier(provider) && matches!(field, "QUEUE" | "RPM" | "TPM" | "CONCURRENCY")
363}
364
365fn is_model_role_name(name: &str) -> bool {
366 let suffix = name
367 .strip_prefix("HARN_LLM_ROLE_")
368 .or_else(|| name.strip_prefix("HARN_LLM_"));
369 let Some(suffix) = suffix else {
370 return false;
371 };
372 ["_MODEL", "_PROVIDER", "_ROUTE_POLICY"]
373 .iter()
374 .find_map(|ending| suffix.strip_suffix(ending))
375 .is_some_and(is_upper_identifier)
376}
377
378fn is_agent_model_option_name(name: &str) -> bool {
382 const SUFFIXES: &[&str] = &[
383 "_EFFORT",
384 "_MODEL",
385 "_MODEL_ROLE",
386 "_PROVIDER",
387 "_REASONING_TASK",
388 "_TOOL_FORMAT",
389 ];
390 let Some(prefix) = SUFFIXES.iter().find_map(|suffix| name.strip_suffix(suffix)) else {
391 return false;
392 };
393 let Some(prefix) = prefix.strip_prefix("HARN_") else {
394 return false;
395 };
396 let role = prefix
397 .strip_prefix("AGENT_")
398 .or_else(|| prefix.strip_prefix("LLM_"))
399 .unwrap_or(prefix);
400 matches!(prefix, "AGENT" | "LLM") || is_upper_identifier(role)
401}
402
403fn is_upper_identifier(value: &str) -> bool {
404 !value.is_empty()
405 && value
406 .bytes()
407 .all(|byte| byte.is_ascii_uppercase() || byte.is_ascii_digit() || byte == b'_')
408 && !value.starts_with('_')
409 && !value.ends_with('_')
410}
411
412fn nearest_registered_name(name: &str) -> Option<String> {
413 registered_names()
414 .iter()
415 .copied()
416 .filter(|candidate| !candidate.ends_with('_'))
417 .map(|candidate| (strsim::levenshtein(name, candidate), candidate))
418 .min_by_key(|(distance, candidate)| (*distance, *candidate))
419 .filter(|(distance, _)| *distance <= 3)
420 .map(|(_, candidate)| candidate.to_string())
421}
422
423#[cfg(test)]
424mod tests {
425 use super::*;
426
427 #[test]
428 fn typo_is_typed_and_suggests_registered_name() {
429 let error = validate_environment([("HARN_LLM_TIMOUT", "30")]).unwrap_err();
430 assert_eq!(
431 error.diagnostics(),
432 &[EnvironmentDiagnostic {
433 code: UNKNOWN_CODE,
434 key: "HARN_LLM_TIMOUT".to_string(),
435 kind: EnvironmentDiagnosticKind::UnknownName {
436 suggestion: Some("HARN_LLM_TIMEOUT".to_string()),
437 },
438 }]
439 );
440 }
441
442 #[test]
443 fn known_and_structured_extension_names_are_accepted() {
444 validate_environment([
445 ("HARN_LLM_TIMEOUT", "30"),
446 ("HARN_EXT_ACME_MODE", "custom"),
447 ("HARN_LLM_ROLE_REVIEW_MODEL", "reviewer"),
448 ("HARN_SECRET_ACME_TOKEN", "credential"),
449 ])
450 .unwrap();
451 }
452
453 #[test]
454 fn malformed_extension_name_is_not_a_prefix_bypass() {
455 let error = validate_environment([("HARN_EXT_", "anything")]).unwrap_err();
456 assert!(matches!(
457 error.diagnostics()[0].kind,
458 EnvironmentDiagnosticKind::UnknownName { .. }
459 ));
460 }
461
462 #[test]
470 fn dynamic_agent_role_options_follow_a_closed_structural_grammar() {
471 for name in [
472 "HARN_AGENT_MODEL",
473 "HARN_LLM_TOOL_FORMAT",
474 "HARN_AGENT_REVIEW_PROVIDER",
475 "HARN_LLM_PLANNER_REASONING_TASK",
476 "HARN_RELEASE_EFFORT",
477 ] {
478 assert!(variable_spec(name).is_some(), "{name}");
479 }
480 for name in [
481 "HARN_AGENT_REVIEW_UNKNOWN",
482 "HARN_LLM_TIMOUT",
483 "HARN_RELEASE_",
484 ] {
485 assert!(variable_spec(name).is_none(), "{name}");
486 }
487 }
488
489 #[test]
490 fn credential_metadata_covers_non_api_oauth_keys() {
491 assert_eq!(
492 variable_spec("HARN_OAUTH_KEY").unwrap().sensitivity,
493 EnvironmentSensitivity::Credential
494 );
495 }
496
497 #[test]
498 fn invalid_known_value_is_rejected_at_startup() {
499 let error = validate_environment([("HARN_LLM_TIMEOUT", "soon")]).unwrap_err();
500 assert_eq!(
501 error.diagnostics()[0].kind,
502 EnvironmentDiagnosticKind::InvalidValue {
503 expected: EnvironmentValueShape::UnsignedInteger
504 }
505 );
506 }
507
508 #[test]
509 fn subtask_placement_uses_the_runtime_owned_closed_vocabulary() {
510 let spec = variable_spec(crate::vm::subtask::PLACEMENT_ENV).unwrap();
511 assert_eq!(
512 spec.value_shape,
513 EnvironmentValueShape::Enumerated(crate::vm::subtask::PLACEMENT_VALUES)
514 );
515 validate_environment([(crate::vm::subtask::PLACEMENT_ENV, "worker")]).unwrap();
516 validate_environment([(crate::vm::subtask::PLACEMENT_ENV, "CURRENT_THREAD")]).unwrap();
517
518 let error =
519 validate_environment([(crate::vm::subtask::PLACEMENT_ENV, "workers")]).unwrap_err();
520 assert_eq!(
521 error.diagnostics()[0].kind,
522 EnvironmentDiagnosticKind::InvalidValue {
523 expected: EnvironmentValueShape::Enumerated(crate::vm::subtask::PLACEMENT_VALUES)
524 }
525 );
526 assert!(error.to_string().contains("`worker`, `current_thread`"));
527 }
528
529 #[test]
530 fn diagnostics_cannot_render_values_even_for_credentials() {
531 let secret = "must-never-appear";
532 let error = validate_environment([("HARN_CLOUD_API_KEZ", secret)]).unwrap_err();
533 let rendered = error.to_string();
534 assert!(rendered.contains("HARN_CLOUD_API_KEZ"));
535 assert!(!rendered.contains(secret));
536 }
537
538 #[test]
539 fn registry_is_sorted_unique_and_contains_metadata() {
540 let names = registered_names();
541 assert!(
542 names.windows(2).all(|pair| pair[0] < pair[1]),
543 "environment registry must remain sorted and unique"
544 );
545 let timeout = variable_spec("HARN_LLM_TIMEOUT").unwrap();
546 assert_eq!(timeout.consumer, EnvironmentConsumer::Runtime);
547 assert_eq!(timeout.value_shape, EnvironmentValueShape::UnsignedInteger);
548 let token = variable_spec("HARN_PACKAGE_REGISTRY_TOKEN").unwrap();
549 assert_eq!(token.sensitivity, EnvironmentSensitivity::Credential);
550 let host_providers = variable_spec(crate::llm_config::HOST_PROVIDERS_CONFIG_ENV).unwrap();
551 assert_eq!(host_providers.consumer, EnvironmentConsumer::Runtime);
552 assert_eq!(
553 host_providers.value_shape,
554 EnvironmentValueShape::OwnerValidated
555 );
556 }
557
558 #[test]
559 fn embedded_runtime_bootstrap_accepts_registered_process_environment() {
560 crate::initialize_runtime().expect("registered process environment");
561 }
562
563 #[test]
564 fn every_compiled_harn_name_is_registered_or_structurally_owned() {
565 let workspace_root = std::path::Path::new(env!("CARGO_MANIFEST_DIR"))
566 .ancestors()
567 .nth(2)
568 .expect("harn-vm lives below workspace/crates");
569 let protocol_symbols = protocol_artifact_symbol_names(workspace_root);
570 let crates_dir = std::path::Path::new(env!("CARGO_MANIFEST_DIR"))
571 .parent()
572 .expect("harn-vm lives below crates");
573 let mut missing = std::collections::BTreeSet::new();
574 for entry in walkdir::WalkDir::new(crates_dir)
575 .into_iter()
576 .filter_map(Result::ok)
577 .filter(|entry| {
578 entry.file_type().is_file()
579 && entry.path().extension().and_then(OsStr::to_str) == Some("rs")
580 && entry
581 .path()
582 .components()
583 .any(|component| component.as_os_str() == "src")
584 && !is_protocol_artifact_projection(entry.path())
585 && entry.file_name() != "environment_registry.rs"
586 })
587 {
588 let source = std::fs::read_to_string(entry.path()).expect("read Rust source");
589 for token in harn_name_tokens(&source) {
590 if variable_spec(token).is_none()
591 && !protocol_symbols.contains(token)
592 && !matches!(token, "HARN_LLM_" | "HARN_LLM_ROLE_" | "HARN_SECRET_")
593 {
594 missing.insert(format!("{}: {token}", entry.path().display()));
595 }
596 }
597 }
598 assert!(
599 missing.is_empty(),
600 "compiled HARN_* names missing from environment_registry_names.txt:\n{}",
601 missing.into_iter().collect::<Vec<_>>().join("\n")
602 );
603 }
604
605 #[test]
606 fn every_harn_script_owned_name_is_registered_or_structurally_owned() {
607 let workspace_root = std::path::Path::new(env!("CARGO_MANIFEST_DIR"))
608 .ancestors()
609 .nth(2)
610 .expect("harn-vm lives below workspace/crates");
611 let source_roots = [
612 "benchmarks",
613 "conformance",
614 "crates",
615 "evals",
616 "examples",
617 "experiments",
618 "perf",
619 "personas",
620 "scripts",
621 "tests",
622 ];
623 let protocol_symbols = protocol_artifact_symbol_names(workspace_root);
624 let mut missing = std::collections::BTreeSet::new();
625 for source_root in source_roots {
626 let source_root = workspace_root.join(source_root);
627 if !source_root.exists() {
628 continue;
629 }
630 for entry in walkdir::WalkDir::new(source_root)
631 .into_iter()
632 .filter_entry(|entry| !is_pruned_reference_directory(entry))
633 .filter_map(Result::ok)
634 .filter(|entry| {
635 entry.file_type().is_file()
636 && entry.path().extension().and_then(OsStr::to_str) == Some("harn")
637 })
638 {
639 let source = std::fs::read_to_string(entry.path()).expect("read Harn source");
640 for token in harn_name_tokens(&source) {
641 if variable_spec(token).is_none()
642 && !protocol_symbols.contains(token)
643 && !matches!(
644 token,
645 "HARN_AGENT"
646 | "HARN_AGENT_"
647 | "HARN_LLM"
648 | "HARN_LLM_"
649 | "HARN_PLANNER"
650 | "HARN_RELEASE"
651 )
652 {
653 missing.insert(format!("{}: {token}", entry.path().display()));
654 }
655 }
656 }
657 }
658 assert!(
659 missing.is_empty(),
660 "Harn-script HARN_* names missing from environment_registry_names.txt:\n{}",
661 missing.into_iter().collect::<Vec<_>>().join("\n")
662 );
663 }
664
665 fn workflow_assigned_names(source: &str) -> std::collections::BTreeSet<String> {
673 let mut names = std::collections::BTreeSet::new();
674 for line in source.lines() {
675 let trimmed = line.trim_start();
676 let candidate = trimmed.strip_prefix("export ").unwrap_or(trimmed);
677 let Some(rest) = candidate.strip_prefix("HARN_") else {
678 continue;
679 };
680 let mut name = String::from("HARN_");
681 let mut terminator = None;
682 for character in rest.chars() {
683 if character.is_ascii_uppercase() || character.is_ascii_digit() || character == '_'
684 {
685 name.push(character);
686 } else {
687 terminator = Some(character);
688 break;
689 }
690 }
691 if name == "HARN_" {
692 continue;
693 }
694 if !matches!(terminator, Some(':') | Some('=')) {
698 continue;
699 }
700 names.insert(name);
701 }
702 names
703 }
704
705 #[test]
706 fn workflow_name_scan_reads_assignments_and_ignores_references() {
707 let source = concat!(
708 " env:\n",
709 " HARN_BUMP_BRANCH: main\n",
710 " run: |\n",
711 " export HARN_BUMP_REFRESH=\"$HARN_BUMP_REFRESH_COMMAND\"\n",
712 " echo \"$HARN_BUMP_TOKEN\"\n",
713 " # HARN_BUMP_NOT_A_KEY: prose\n",
714 );
715 let names = workflow_assigned_names(source);
716 assert!(names.contains("HARN_BUMP_BRANCH"), "reads a YAML env key");
717 assert!(names.contains("HARN_BUMP_REFRESH"), "reads a shell export");
718 assert!(
719 !names.contains("HARN_BUMP_REFRESH_COMMAND"),
720 "a reference on a right-hand side is not an assignment"
721 );
722 assert!(
723 !names.contains("HARN_BUMP_TOKEN"),
724 "an echoed reference is not an assignment"
725 );
726 assert!(
727 !names.contains("HARN_BUMP_NOT_A_KEY"),
728 "a comment is not an assignment"
729 );
730 }
731
732 #[test]
737 fn every_workflow_assigned_name_is_registered() {
738 let workspace_root = std::path::Path::new(env!("CARGO_MANIFEST_DIR"))
739 .ancestors()
740 .nth(2)
741 .expect("harn-vm lives below workspace/crates");
742 let workflows = workspace_root.join(".github/workflows");
743 assert!(
744 workflows.is_dir(),
745 "no workflow directory at {}; this gate would pass by seeing nothing",
746 workflows.display()
747 );
748 let mut scanned = 0usize;
749 let mut missing = std::collections::BTreeSet::new();
750 for entry in walkdir::WalkDir::new(&workflows)
751 .into_iter()
752 .filter_map(Result::ok)
753 .filter(|entry| {
754 entry.file_type().is_file()
755 && matches!(
756 entry.path().extension().and_then(OsStr::to_str),
757 Some("yml") | Some("yaml")
758 )
759 })
760 {
761 let source = std::fs::read_to_string(entry.path()).expect("read workflow");
762 scanned += 1;
763 for name in workflow_assigned_names(&source) {
764 if variable_spec(&name).is_none() {
765 missing.insert(format!("{}: {name}", entry.path().display()));
766 }
767 }
768 }
769 assert!(
770 scanned > 0,
771 "scanned no workflow files; the gate cannot see anything"
772 );
773 assert!(
774 missing.is_empty(),
775 "workflow-assigned HARN_* names missing from \
776 environment_registry_names.txt:\n{}",
777 missing.into_iter().collect::<Vec<_>>().join("\n")
778 );
779 }
780
781 const UNREAD_NAME_ALLOWLIST: &[(&str, &str)] = &[];
787
788 fn is_pruned_reference_directory(entry: &walkdir::DirEntry) -> bool {
793 if !entry.file_type().is_dir() {
794 return false;
795 }
796 let Some(name) = entry.file_name().to_str() else {
797 return true;
798 };
799 name.starts_with(".target")
800 || matches!(
801 name,
802 ".build"
803 | ".burin"
804 | ".git"
805 | ".harn"
806 | ".harn-runs"
807 | ".harn-toolchain-cache"
808 | ".venv"
809 | ".worktrees"
810 | "__pycache__"
811 | "changelog"
812 | "dist"
813 | "node_modules"
814 | "pkg"
815 | "target"
816 )
817 }
818
819 #[test]
820 fn generated_harn_state_is_not_an_environment_contract_owner() {
821 let root = tempfile::tempdir().expect("temporary workspace");
822 let state = root.path().join(".harn");
823 std::fs::create_dir(&state).expect("create generated Harn state");
824 let entry = walkdir::WalkDir::new(root.path())
825 .into_iter()
826 .filter_map(Result::ok)
827 .find(|entry| entry.path() == state)
828 .expect("walk generated Harn state");
829
830 assert!(is_pruned_reference_directory(&entry));
831 }
832
833 fn is_protocol_artifact_projection(path: &std::path::Path) -> bool {
837 path.components().any(|component| {
838 matches!(
839 component.as_os_str().to_str(),
840 Some("dump_protocol_artifacts" | "protocol-artifacts")
841 )
842 })
843 }
844
845 fn is_reference_source(path: &std::path::Path) -> bool {
849 if path.extension().and_then(OsStr::to_str) == Some("md")
850 || is_protocol_artifact_projection(path)
851 {
852 return false;
853 }
854 !matches!(
855 path.file_name().and_then(OsStr::to_str),
856 Some("environment_registry.rs" | "environment_registry_names.txt")
857 )
858 }
859
860 fn protocol_artifact_symbol_names(
864 workspace_root: &std::path::Path,
865 ) -> std::collections::BTreeSet<String> {
866 let path = workspace_root.join("spec/protocol-artifacts/harn-protocol.rs");
867 let source = std::fs::read_to_string(&path)
868 .unwrap_or_else(|error| panic!("read {}: {error}", path.display()));
869 bounded_harn_tokens(&source).map(str::to_string).collect()
870 }
871
872 #[test]
873 fn protocol_artifact_names_are_not_environment_references() {
874 let workspace_root = std::path::Path::new(env!("CARGO_MANIFEST_DIR"))
875 .ancestors()
876 .nth(2)
877 .expect("harn-vm lives below workspace/crates");
878 let protocol_symbols = protocol_artifact_symbol_names(workspace_root);
879 let generator = std::path::Path::new(
880 "crates/harn-cli/src/commands/dump_protocol_artifacts/typescript.rs",
881 );
882 let projection = std::path::Path::new("spec/protocol-artifacts/harn-protocol.ts");
883 let runtime = std::path::Path::new("crates/harn-vm/src/llm/call.rs");
884
885 assert!(is_protocol_artifact_projection(generator));
886 assert!(is_protocol_artifact_projection(projection));
887 assert!(!is_reference_source(generator));
888 assert!(!is_reference_source(projection));
889 assert!(is_reference_source(runtime));
890 assert!(variable_spec("HARN_AGENT_EVENT_KINDS").is_none());
891 assert!(variable_spec("HARN_WORKER_STATUSES").is_none());
892 assert!(protocol_symbols.contains("HARN_PROTOCOL_ARTIFACT_VERSION"));
893 assert!(protocol_symbols.contains("HARN_AGENT_EVENT_KINDS"));
894 assert!(!protocol_symbols.contains("HARN_ACP_VERBOSE"));
895 }
896
897 #[derive(Clone, Copy, Debug, Eq, PartialEq)]
898 enum EnvironmentReferenceSyntax {
899 StringLiteral,
900 ShellLike,
901 }
902
903 fn environment_reference_syntax(path: &std::path::Path) -> EnvironmentReferenceSyntax {
904 let file_name = path.file_name().and_then(OsStr::to_str).unwrap_or_default();
905 let extension = path.extension().and_then(OsStr::to_str).unwrap_or_default();
906 if matches!(file_name, "Makefile" | "GNUmakefile")
907 || file_name.starts_with(".env")
908 || matches!(
909 extension,
910 "bash" | "bat" | "cmd" | "fish" | "mk" | "ps1" | "sh" | "yaml" | "yml" | "zsh"
911 )
912 {
913 EnvironmentReferenceSyntax::ShellLike
914 } else {
915 EnvironmentReferenceSyntax::StringLiteral
916 }
917 }
918
919 fn environment_reference_tokens<'a>(path: &std::path::Path, source: &'a str) -> Vec<&'a str> {
920 match environment_reference_syntax(path) {
921 EnvironmentReferenceSyntax::StringLiteral => harn_name_tokens(source).collect(),
922 EnvironmentReferenceSyntax::ShellLike => bounded_harn_tokens(source).collect(),
923 }
924 }
925
926 #[expect(
930 clippy::string_slice,
931 reason = "start/end bound an ASCII HARN_* token found by match_indices"
932 )]
933 fn bounded_harn_tokens(source: &str) -> impl Iterator<Item = &str> {
934 let bytes = source.as_bytes();
935 source.match_indices("HARN_").filter_map(move |(start, _)| {
936 if start > 0 {
937 let previous = bytes[start - 1];
938 if previous.is_ascii_alphanumeric() || previous == b'_' {
939 return None;
940 }
941 }
942 let mut end = start + "HARN_".len();
943 while end < bytes.len()
944 && (bytes[end].is_ascii_uppercase()
945 || bytes[end].is_ascii_digit()
946 || bytes[end] == b'_')
947 {
948 end += 1;
949 }
950 Some(&source[start..end])
951 })
952 }
953
954 #[test]
959 fn every_registered_name_has_a_non_projection_source_reference() {
960 let workspace_root = std::path::Path::new(env!("CARGO_MANIFEST_DIR"))
961 .ancestors()
962 .nth(2)
963 .expect("harn-vm lives below workspace/crates");
964 let mut referenced = std::collections::BTreeSet::new();
965 let protocol_symbols = protocol_artifact_symbol_names(workspace_root);
966 for entry in walkdir::WalkDir::new(workspace_root)
967 .into_iter()
968 .filter_entry(|entry| !is_pruned_reference_directory(entry))
969 .filter_map(Result::ok)
970 .filter(|entry| entry.file_type().is_file() && is_reference_source(entry.path()))
971 {
972 let Ok(source) = std::fs::read_to_string(entry.path()) else {
973 continue;
974 };
975 for token in environment_reference_tokens(entry.path(), &source) {
976 if !protocol_symbols.contains(token) {
977 referenced.insert(token.to_string());
978 }
979 }
980 }
981
982 let allowed: std::collections::BTreeSet<&str> = UNREAD_NAME_ALLOWLIST
983 .iter()
984 .map(|(name, _)| *name)
985 .collect();
986 let unregistered_allowlist = allowed
987 .iter()
988 .copied()
989 .filter(|name| registered_names().binary_search(name).is_err())
990 .collect::<Vec<_>>();
991 assert!(
992 unregistered_allowlist.is_empty(),
993 "allowlisted names are not in environment_registry_names.txt:\n{}",
994 unregistered_allowlist.join("\n")
995 );
996 let owned_allowlist = allowed
997 .iter()
998 .copied()
999 .filter(|name| referenced.contains(*name))
1000 .collect::<Vec<_>>();
1001 assert!(
1002 owned_allowlist.is_empty(),
1003 "allowlisted names now have source owners; drop them from UNREAD_NAME_ALLOWLIST:\n{}",
1004 owned_allowlist.join("\n")
1005 );
1006
1007 let unread = registered_names()
1008 .iter()
1009 .copied()
1010 .filter(|name| !referenced.contains(*name) && !allowed.contains(name))
1011 .collect::<Vec<_>>();
1012 assert!(
1013 unread.is_empty(),
1014 "registered names have no source owner; delete them from \
1015 environment_registry_names.txt or allowlist them with a reason:\n{}",
1016 unread.join("\n")
1017 );
1018 }
1019
1020 #[test]
1021 fn environment_reference_scan_dispatches_by_source_syntax() {
1022 let compiled = concat!(
1023 "const HARN_AGENT_EVENT_KINDS: &[&str] = &[];\n",
1024 "const ENV: &str = \"HARN_REAL_ENVIRONMENT_KNOB\";\n",
1025 );
1026 assert_eq!(
1027 environment_reference_tokens(std::path::Path::new("runtime.rs"), compiled),
1028 vec!["HARN_REAL_ENVIRONMENT_KNOB"]
1029 );
1030 assert_eq!(
1031 environment_reference_tokens(
1032 std::path::Path::new("bench.sh"),
1033 "cache=${HARN_BENCH_CACHE_DIR:-target}\n",
1034 ),
1035 vec!["HARN_BENCH_CACHE_DIR"]
1036 );
1037 assert_eq!(
1038 environment_reference_tokens(
1039 std::path::Path::new("Makefile"),
1040 "HARN_BIN_ASSIGN = harn_bin\n",
1041 ),
1042 vec!["HARN_BIN_ASSIGN"]
1043 );
1044 }
1045
1046 #[expect(
1047 clippy::string_slice,
1048 reason = "start/end bound an ASCII HARN_* token found by match_indices"
1049 )]
1050 fn harn_name_tokens(source: &str) -> impl Iterator<Item = &str> {
1051 source.match_indices("\"HARN_").filter_map(|(quote, _)| {
1052 let start = quote + 1;
1053 let bytes = source.as_bytes();
1054 let mut end = start + "HARN_".len();
1055 while end < bytes.len()
1056 && (bytes[end].is_ascii_uppercase()
1057 || bytes[end].is_ascii_digit()
1058 || bytes[end] == b'_')
1059 {
1060 end += 1;
1061 }
1062 (end > start + "HARN_".len()).then(|| &source[start..end])
1063 })
1064 }
1065}