@holmes-lab/holmes-kit 0.19.0 → 0.19.3
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +120 -0
- package/dist/.build-id +1 -1
- package/dist/holmes/cli/agents.d.ts +22 -0
- package/dist/holmes/cli/agents.js +76 -1
- package/dist/holmes/cli/approve.js +6 -1
- package/dist/holmes/cli/doctor.d.ts +51 -1
- package/dist/holmes/cli/doctor.js +211 -36
- package/dist/holmes/cli/index.js +7 -1
- package/dist/holmes/cli/init.js +12 -0
- package/dist/holmes/cli/native-deps.d.ts +65 -0
- package/dist/holmes/cli/native-deps.js +131 -0
- package/dist/holmes/cpg/cycle-observation.d.ts +65 -0
- package/dist/holmes/cpg/cycle-observation.js +146 -0
- package/dist/holmes/governance/approval-queue.d.ts +23 -4
- package/dist/holmes/governance/approval-queue.js +44 -6
- package/dist/holmes/hooks/stop.d.ts +15 -0
- package/dist/holmes/hooks/stop.js +46 -3
- package/dist/holmes/mcp/handlers.d.ts +2 -0
- package/dist/holmes/mcp/handlers.js +29 -2
- package/dist/holmes/mcp/maintenance-analyze.d.ts +37 -0
- package/dist/holmes/mcp/maintenance-analyze.js +73 -1
- package/dist/holmes/mcp/maintenance-evidence.d.ts +41 -0
- package/dist/holmes/mcp/maintenance-evidence.js +71 -4
- package/dist/holmes/project/install-scripts-policy.d.ts +76 -0
- package/dist/holmes/project/install-scripts-policy.js +131 -0
- package/dist/holmes/project/npx-bin.d.ts +6 -0
- package/dist/holmes/project/npx-bin.js +10 -0
- package/dist/holmes/review/failed-test-names.d.ts +19 -0
- package/dist/holmes/review/failed-test-names.js +43 -0
- package/dist/holmes/review/run-replay.d.ts +23 -0
- package/dist/holmes/review/run-replay.js +30 -0
- package/dist/holmes/review/test-runner.d.ts +27 -0
- package/dist/holmes/review/test-runner.js +59 -3
- package/docs/install-guide.md +54 -5
- package/package.json +4 -1
package/CHANGELOG.md
CHANGED
|
@@ -5,6 +5,126 @@ All notable changes to this project will be documented in this file.
|
|
|
5
5
|
The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
|
|
6
6
|
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
|
|
7
7
|
<!-- @implements A-SPEC-209 -->
|
|
8
|
+
## [0.19.3] - 2026-09-10
|
|
9
|
+
|
|
10
|
+
`doctor` told every third-party install its SQLite binding was missing while it sat on disk. No API change.
|
|
11
|
+
|
|
12
|
+
### Fixed
|
|
13
|
+
- **A false FAIL on the default install layout** (REQ-583). `gatherNativeEvidence` located
|
|
14
|
+
`better-sqlite3` by joining `packageRoot/node_modules/better-sqlite3`, and npm HOISTS
|
|
15
|
+
dependencies to the project root — so for a normal install at
|
|
16
|
+
`<proj>/node_modules/@holmes-lab/holmes-kit` the joined path never existed. Measured against the
|
|
17
|
+
published 0.19.2: the binding was present, `require('better-sqlite3')` executed a `:memory:`
|
|
18
|
+
query, and `doctor` still reported `no binary`. Hoisting is npm's default, which made this the
|
|
19
|
+
default experience of a correct install. The package directory is now RESOLVED
|
|
20
|
+
(`require.resolve('<name>/package.json', { paths: [root] })`) — Node's own rule, so hoisted,
|
|
21
|
+
nested and linked layouts all answer correctly, and a genuinely absent dependency still fails.
|
|
22
|
+
Its sibling — the `node version` check reading `engines.node` — carried the same join and moved
|
|
23
|
+
with it; it had been reporting `engines.node=(unknown)` for the same reason.
|
|
24
|
+
- **The release gate ran the suite in the operator's environment** (REQ-582). `npm publish`
|
|
25
|
+
refused a green tree: `verify-release.js` called `npm test` with the shell's `HOLMES_APPROVAL`,
|
|
26
|
+
which the spawned-hook parity test (A-SPEC-338) legitimately reacts to — the spawned child
|
|
27
|
+
inherits the token while its in-process twin is handed `{approval: undefined}`. The gate now runs
|
|
28
|
+
under `cleanTestEnv`, the same scrub `test_run` uses (borrowed, never copied), and refuses to
|
|
29
|
+
proceed unscrubbed if it cannot load it. It also names the failing tests instead of saying only
|
|
30
|
+
"red" — finding out which test had failed previously cost two extra full-suite runs.
|
|
31
|
+
|
|
32
|
+
- **A live session could block its own publish** (REQ-583). The `APPEND_ONLY_GOVERNANCE` list that
|
|
33
|
+
tells the release gate which dirty files are governance churn rather than unshipped code held
|
|
34
|
+
only the approval queue and provenance — while REQ-564/566/569/578 added `session-context`,
|
|
35
|
+
`impact-advisories`, `anchor-density` and `cycle-observations`, and `test-outcomes` was promoted
|
|
36
|
+
to tracking. Measured: `npm publish` refused because one `session-context` line was appended
|
|
37
|
+
between two `git status` snapshots inside the tree-untouched assertion, and the same list guards
|
|
38
|
+
the dirty-tree check a user meets first. The list now follows the ledgers, and a test pins that
|
|
39
|
+
nothing which ships (`src/`, `dist/`, the manifest, specs) is ever excused.
|
|
40
|
+
|
|
41
|
+
### Verified against the published package
|
|
42
|
+
- Installed `@holmes-lab/holmes-kit@0.19.2` from the registry into a clean project, applied this
|
|
43
|
+
fix, and re-diagnosed from an unrelated working directory: **0 FAIL**, three-harness handshakes
|
|
44
|
+
(claude / codex / antigravity) all PASS on the version-pinned npx wiring, `target wiring — 3
|
|
45
|
+
holmes hook(s) resolve to this install`, and `better-sqlite3 — loads and executes against
|
|
46
|
+
:memory:`.
|
|
47
|
+
|
|
48
|
+
## [0.19.2] - 2026-09-10
|
|
49
|
+
|
|
50
|
+
Wiring diagnosis no longer depends on where you run it. No API change.
|
|
51
|
+
|
|
52
|
+
### Fixed
|
|
53
|
+
- **`doctor` gave opposite verdicts for the same wiring depending on the directory it was run
|
|
54
|
+
from** (REQ-581). A relative executable path in a wiring file is resolved by Node against
|
|
55
|
+
`process.cwd()`, and a relative path is the CONVENTION here — `.mcp.json` has always carried
|
|
56
|
+
`bin/holmes-mcp.js`. Measured: the same project under the same `--target` read
|
|
57
|
+
`PASS codex wiring` from the repository root and `FAIL — points at a missing file` from `/tmp`,
|
|
58
|
+
with the fix line telling the operator to re-init a wiring that was never wrong. All four
|
|
59
|
+
affected checks (`mcp wiring spawn`, `codex wiring`, and the `wiringHandshakeChecks` that cover
|
|
60
|
+
all three harnesses) now resolve against the TARGET project, and the handshake launches the
|
|
61
|
+
server FROM that project — the way a harness actually starts it. Absolute wirings (antigravity's)
|
|
62
|
+
are returned untouched, and a path that is genuinely missing still fails. The message keeps the
|
|
63
|
+
string exactly as written in the wiring file: what an operator must fix is that string, not our
|
|
64
|
+
arithmetic.
|
|
65
|
+
- **The same wiring could read PASS from one check and FAIL from its sibling.** Measured on a
|
|
66
|
+
third-party tarball install: `claude wiring handshake` passed while `mcp wiring spawn` failed on
|
|
67
|
+
the identical `npx` command, because only one of the two spawn paths had been moved. Both now
|
|
68
|
+
launch from the target.
|
|
69
|
+
|
|
70
|
+
### Verified on a third-party install (macOS, npm 11.17.0, Node 26.5.1)
|
|
71
|
+
- Tarball installed into a clean project: `better-sqlite3` binding present, `:memory:` query ok.
|
|
72
|
+
- With install scripts blocked (`--ignore-scripts`, the npm 12 condition): the CLI still starts,
|
|
73
|
+
`init` still wires, the MCP server still answers — and `doctor` names the cause and prints a
|
|
74
|
+
recovery command that works verbatim (`npm rebuild better-sqlite3 --foreground-scripts`).
|
|
75
|
+
- Diagnosed from an unrelated working directory with the consumer's own CLI: claude / codex /
|
|
76
|
+
antigravity handshakes 3/3 PASS, `target wiring — 3 holmes hook(s) resolve to this install`.
|
|
77
|
+
|
|
78
|
+
### Corrected from 0.19.1's notes
|
|
79
|
+
- 0.19.1 recorded that `codexMcpBlock` "writes Windows backslash paths into `config.toml`
|
|
80
|
+
unescaped". Re-measured: it does not. `tomlStr` escapes `\` correctly and the write→read
|
|
81
|
+
round-trip returns `C:\Users\name\proj\bin\holmes-mcp.js` byte-identical (pure functions, so
|
|
82
|
+
the result is platform-independent). The `doctor-wiring-handshake` failure seen on a Windows
|
|
83
|
+
checkout is therefore unexplained by escaping — and is a candidate instance of the cwd defect
|
|
84
|
+
fixed above, which is a HYPOTHESIS for the next Windows run, not a conclusion.
|
|
85
|
+
|
|
86
|
+
## [0.19.1] - 2026-09-10
|
|
87
|
+
|
|
88
|
+
Windows install hardening, measured on Windows 11 / npm 12.0.1 / Node 24.19.0. No API change.
|
|
89
|
+
|
|
90
|
+
### Fixed
|
|
91
|
+
- **npm 12 skipped `better-sqlite3`'s install script and nothing said so** (REQ-579). npm ≥ 12
|
|
92
|
+
(and npm 11.19+) blocks dependency install scripts unless the root `package.json` approves
|
|
93
|
+
them; `npm ci` then succeeds with no `better_sqlite3.node`, and every SQLite-backed feature
|
|
94
|
+
fails on first use. `package.json` now declares `allowScripts: { "better-sqlite3@12.11.1": true }`
|
|
95
|
+
— the ONE dependency that needs its script, pinned to the lockfile version so a bump forces a
|
|
96
|
+
fresh review (a test goes red until the pin is updated). The 8 tree-sitter packages are
|
|
97
|
+
deliberately not approved: measured, they load from their shipped `prebuilds/` with the script
|
|
98
|
+
blocked. Older npm 11 ignores the field.
|
|
99
|
+
- **doctor misdiagnosed every missing SQLite binary as an ABI mismatch** (REQ-580). The
|
|
100
|
+
`better-sqlite3` check now judges from evidence — binary present or not, the load error's
|
|
101
|
+
wording, the npm major, the governing `allowScripts`, the install layout (repository / project
|
|
102
|
+
dependency / global / npx), and on Windows the toolchain and a spaced path — and names the
|
|
103
|
+
cause: `scripts-blocked`, `abi-mismatch`, `build-failed`, or `unknown` with the raw error. A
|
|
104
|
+
prebuilt-download failure is stated as unobservable, never asserted. The fix line is the exact
|
|
105
|
+
command for that layout; for a global install it targets the dependency
|
|
106
|
+
(`npm rebuild -g better-sqlite3 --foreground-scripts --allow-scripts=better-sqlite3`), because
|
|
107
|
+
`rebuild -g @holmes-lab/holmes-kit` dies `EEXIST` re-linking the bin under npm 12 (measured).
|
|
108
|
+
- **doctor's grammar check only resolved packages.** It now parses a source with all 8 grammars
|
|
109
|
+
(typescript, tsx, python, c-sharp, java, go, rust, cpp) in a fresh child process — in-process
|
|
110
|
+
parsing is unreliable after another module registry has loaded tree-sitter's native addon.
|
|
111
|
+
- **Windows recovery commands are runnable.** npm commands are emitted as `npm.cmd` on win32
|
|
112
|
+
(`npmBin`, the twin of `npxBin`), and when a `Restricted`/`AllSigned` PowerShell execution policy
|
|
113
|
+
is observed the verdict says the `.ps1` shims are blocked and the `.cmd` ones are not.
|
|
114
|
+
|
|
115
|
+
### Docs
|
|
116
|
+
- `docs/install-guide.md`: npm 12 section, per-layout recovery table, PowerShell policy note,
|
|
117
|
+
build-failed evidence. `docs/windows-test-plan-2026-09.md`: items E/F/G updated with the
|
|
118
|
+
2026-09-09/10 measurements and the known Windows-only baseline failures.
|
|
119
|
+
|
|
120
|
+
### Known, not fixed (recorded honestly)
|
|
121
|
+
- macOS/Linux were not exercised on hardware in this release; their behaviour is covered by
|
|
122
|
+
platform-injected unit tests and the three-harness parity suite only.
|
|
123
|
+
- `codexMcpBlock` writes Windows backslash paths into `config.toml` unescaped
|
|
124
|
+
(`doctor-wiring-handshake` fails on a Windows checkout); doctor's `global prefix` check reports a
|
|
125
|
+
protected `C:\Program Files\nodejs` as writable. Both are pre-existing and tracked as follow-ups.
|
|
126
|
+
**(The escaping half of this was re-measured in 0.19.2 and does not reproduce — see that entry.)**
|
|
127
|
+
|
|
8
128
|
## [0.19.0] - 2026-09-08
|
|
9
129
|
|
|
10
130
|
### Added
|
package/dist/.build-id
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
|
|
1
|
+
fb773a8-mtug045o
|
|
@@ -62,6 +62,28 @@ export declare function antigravityHookCommand(scriptPath: string): string;
|
|
|
62
62
|
/** win32-only, best-effort: a spaced path becomes its 8.3 short name (no spaces); anything else — or
|
|
63
63
|
* any lookup failure — returns the input unchanged. */
|
|
64
64
|
export declare function shortPathIfSpaced(p: string): string;
|
|
65
|
+
export declare const MANAGED_BEGIN = "<!-- holmes-kit:managed:begin -->";
|
|
66
|
+
export declare const MANAGED_END = "<!-- holmes-kit:managed:end -->";
|
|
67
|
+
/**
|
|
68
|
+
* @implements A-SPEC-576.2
|
|
69
|
+
* The regeneration that keeps what it did not write.
|
|
70
|
+
*
|
|
71
|
+
* Measured cause: `init` rewrote AGENTS.md from the template, deleting the ADR-018 parity rule a
|
|
72
|
+
* commit had added the day before — and said nothing, so the loss was found weeks later by grep.
|
|
73
|
+
* Generated text now lives inside a MANAGED BLOCK; everything outside it belongs to the person
|
|
74
|
+
* whose repository this is.
|
|
75
|
+
*
|
|
76
|
+
* PURE, and it takes the whole file rather than a path: the caller reads `before` during init's
|
|
77
|
+
* COMPUTE phase, so a dry-run predicts the same bytes the real run writes (A-SPEC-190 §9).
|
|
78
|
+
*
|
|
79
|
+
* `preserved` is the report, not a courtesy. A run that moved something and returned an empty list
|
|
80
|
+
* would be the same silent loss in a new costume.
|
|
81
|
+
*/
|
|
82
|
+
export declare function mergeAgentsMd(existing: string | null, generated: string): {
|
|
83
|
+
content: string;
|
|
84
|
+
preserved: string[];
|
|
85
|
+
note: string | null;
|
|
86
|
+
};
|
|
65
87
|
/**
|
|
66
88
|
* 이 하네스에 써야 할 파일들. **쓰지는 않는다** — 무엇을 쓸지만 말한다.
|
|
67
89
|
*
|
|
@@ -33,9 +33,10 @@ var __importStar = (this && this.__importStar) || (function () {
|
|
|
33
33
|
};
|
|
34
34
|
})();
|
|
35
35
|
Object.defineProperty(exports, "__esModule", { value: true });
|
|
36
|
-
exports.CODEX_MARKETPLACE = exports.CODEX_PLUGIN_DIR = exports.HARNESS_ENFORCES = exports.AGENTS = void 0;
|
|
36
|
+
exports.MANAGED_END = exports.MANAGED_BEGIN = exports.CODEX_MARKETPLACE = exports.CODEX_PLUGIN_DIR = exports.HARNESS_ENFORCES = exports.AGENTS = void 0;
|
|
37
37
|
exports.antigravityHookCommand = antigravityHookCommand;
|
|
38
38
|
exports.shortPathIfSpaced = shortPathIfSpaced;
|
|
39
|
+
exports.mergeAgentsMd = mergeAgentsMd;
|
|
39
40
|
exports.agentFiles = agentFiles;
|
|
40
41
|
exports.agentLinks = agentLinks;
|
|
41
42
|
// @implements A-SPEC-442
|
|
@@ -197,6 +198,80 @@ ${enforced
|
|
|
197
198
|
? `Gates are **enforced**. Tool calls that write un-anchored code or target unapproved specifications will be denied. Unresolved critical findings block completion.`
|
|
198
199
|
: `This harness does not have measured **hook enforcement points**. Holmes-Kit provides tools and guidance here without active gate blocking.`}
|
|
199
200
|
`;
|
|
201
|
+
exports.MANAGED_BEGIN = '<!-- holmes-kit:managed:begin -->';
|
|
202
|
+
exports.MANAGED_END = '<!-- holmes-kit:managed:end -->';
|
|
203
|
+
const PRESERVED_NOTE = '<!-- kept from this file by holmes-kit init — holmes-kit does not write below this line -->';
|
|
204
|
+
/**
|
|
205
|
+
* @implements A-SPEC-576.2
|
|
206
|
+
* The regeneration that keeps what it did not write.
|
|
207
|
+
*
|
|
208
|
+
* Measured cause: `init` rewrote AGENTS.md from the template, deleting the ADR-018 parity rule a
|
|
209
|
+
* commit had added the day before — and said nothing, so the loss was found weeks later by grep.
|
|
210
|
+
* Generated text now lives inside a MANAGED BLOCK; everything outside it belongs to the person
|
|
211
|
+
* whose repository this is.
|
|
212
|
+
*
|
|
213
|
+
* PURE, and it takes the whole file rather than a path: the caller reads `before` during init's
|
|
214
|
+
* COMPUTE phase, so a dry-run predicts the same bytes the real run writes (A-SPEC-190 §9).
|
|
215
|
+
*
|
|
216
|
+
* `preserved` is the report, not a courtesy. A run that moved something and returned an empty list
|
|
217
|
+
* would be the same silent loss in a new costume.
|
|
218
|
+
*/
|
|
219
|
+
function mergeAgentsMd(existing, generated) {
|
|
220
|
+
const block = exports.MANAGED_BEGIN + '\n' + (generated.endsWith('\n') ? generated : generated + '\n') + exports.MANAGED_END + '\n';
|
|
221
|
+
if (existing === null || existing === '')
|
|
222
|
+
return { content: block, preserved: [], note: null };
|
|
223
|
+
const b = existing.indexOf(exports.MANAGED_BEGIN);
|
|
224
|
+
const e = existing.indexOf(exports.MANAGED_END);
|
|
225
|
+
if (b !== -1 && e > b) {
|
|
226
|
+
// The ordinary case once a file has been through here: swap the block, and do not touch a
|
|
227
|
+
// single byte on either side of it.
|
|
228
|
+
const body = generated.endsWith('\n') ? generated : generated + '\n';
|
|
229
|
+
return {
|
|
230
|
+
content: existing.slice(0, b) + exports.MANAGED_BEGIN + '\n' + body + exports.MANAGED_END + existing.slice(e + exports.MANAGED_END.length),
|
|
231
|
+
preserved: [],
|
|
232
|
+
note: null,
|
|
233
|
+
};
|
|
234
|
+
}
|
|
235
|
+
// LEGACY: a file written before the markers existed — every AGENTS.md in the field today. Its
|
|
236
|
+
// lines are compared to the generated ones, and whatever the generator never wrote is carried
|
|
237
|
+
// below the block. Line terminators are kept as read, so a CRLF checkout stays CRLF.
|
|
238
|
+
const known = new Set(generated.split(/\r?\n/).map((l) => l.trim()).filter((l) => l !== ''));
|
|
239
|
+
const lines = existing.split(/(?<=\n)/);
|
|
240
|
+
const runs = [];
|
|
241
|
+
let run = [];
|
|
242
|
+
let pendingBlanks = [];
|
|
243
|
+
for (const line of lines) {
|
|
244
|
+
const t = line.trim();
|
|
245
|
+
if (t === '') {
|
|
246
|
+
if (run.length > 0)
|
|
247
|
+
pendingBlanks.push(line);
|
|
248
|
+
continue;
|
|
249
|
+
}
|
|
250
|
+
if (known.has(t)) {
|
|
251
|
+
run = closeRun(runs, run);
|
|
252
|
+
pendingBlanks = [];
|
|
253
|
+
continue;
|
|
254
|
+
}
|
|
255
|
+
// A blank line INSIDE a run is part of it; blanks trailing a run are not.
|
|
256
|
+
run.push(...pendingBlanks, line);
|
|
257
|
+
pendingBlanks = [];
|
|
258
|
+
}
|
|
259
|
+
closeRun(runs, run);
|
|
260
|
+
if (runs.length === 0)
|
|
261
|
+
return { content: block, preserved: [], note: null };
|
|
262
|
+
const preserved = runs.map((r) => (r.find((l) => l.trim() !== '') ?? '').trim());
|
|
263
|
+
return {
|
|
264
|
+
content: block + '\n' + PRESERVED_NOTE + '\n\n' + runs.map((r) => r.join('')).join('\n'),
|
|
265
|
+
preserved,
|
|
266
|
+
note: `kept ${runs.length} section(s) this generator did not write`,
|
|
267
|
+
};
|
|
268
|
+
}
|
|
269
|
+
/** Close the current run into `runs` (dropping an empty one) and hand back a fresh one. */
|
|
270
|
+
function closeRun(runs, run) {
|
|
271
|
+
if (run.length > 0)
|
|
272
|
+
runs.push(run);
|
|
273
|
+
return [];
|
|
274
|
+
}
|
|
200
275
|
/**
|
|
201
276
|
* 이 하네스에 써야 할 파일들. **쓰지는 않는다** — 무엇을 쓸지만 말한다.
|
|
202
277
|
*
|
|
@@ -78,7 +78,12 @@ const DEFAULT_TTL_MINUTES = 30;
|
|
|
78
78
|
const MAX_TTL_MINUTES = 7 * 24 * 60;
|
|
79
79
|
const DEFAULT_RATIONALE = 'approved via holmes-kit approve';
|
|
80
80
|
const findPending = (root, id) => {
|
|
81
|
-
|
|
81
|
+
// @implements A-SPEC-576.1 — RESOLUTION reads everything; the human's LIST reads decisions only.
|
|
82
|
+
// The inbox filter hides gate refusals from the screen, but `approve <id>` must still find one:
|
|
83
|
+
// the fallback below only holds entries written after the REQ-563 split, so anything refused
|
|
84
|
+
// before it lives in the queue alone. Filtering here would cut the out-of-band approval path the
|
|
85
|
+
// deny message itself promises.
|
|
86
|
+
const state = (0, approval_queue_1.readQueue)(root, { includeAllKinds: true });
|
|
82
87
|
const pending = state.pending.find((p) => p.id === id);
|
|
83
88
|
if (pending)
|
|
84
89
|
return { entry: pending, state };
|
|
@@ -1,3 +1,4 @@
|
|
|
1
|
+
import { NativeEvidence } from './native-deps';
|
|
1
2
|
import { type SemanticTier } from '../semantic/tier';
|
|
2
3
|
/**
|
|
3
4
|
* `holmes-kit doctor` — install diagnostics.
|
|
@@ -82,6 +83,43 @@ export declare function prefixVerdict(input: PrefixProbe): {
|
|
|
82
83
|
detail: string;
|
|
83
84
|
fix?: string;
|
|
84
85
|
};
|
|
86
|
+
/**
|
|
87
|
+
* @implements A-SPEC-580
|
|
88
|
+
* Parse a trivial source with every grammar, in a child node whose module resolution starts at
|
|
89
|
+
* `packageRoot`. `parsed` counts typescript and tsx separately (8 for the 7 grammar packages).
|
|
90
|
+
*/
|
|
91
|
+
export declare function grammarProbe(packageRoot: string): {
|
|
92
|
+
parsed: number;
|
|
93
|
+
failed: string[];
|
|
94
|
+
error?: string;
|
|
95
|
+
};
|
|
96
|
+
/**
|
|
97
|
+
* @implements A-SPEC-580
|
|
98
|
+
* Everything observable about the better-sqlite3 install, gathered for `nativeVerdict`. Each probe
|
|
99
|
+
* is independent and optional: a spawn that fails leaves its field `undefined` (not observed), and
|
|
100
|
+
* the verdict says so rather than treating silence as evidence.
|
|
101
|
+
*
|
|
102
|
+
* Which package.json holds the approval policy depends on the layout: this checkout's own for a
|
|
103
|
+
* repository run, the PROJECT's (three levels up from node_modules/@holmes-lab/holmes-kit) for a
|
|
104
|
+
* local dependency. A global or npx install has no project package.json that npm consults, so its
|
|
105
|
+
* coverage is `uncovered` by construction and the remedy is the per-command `--allow-scripts`.
|
|
106
|
+
*/
|
|
107
|
+
/**
|
|
108
|
+
* Where an installed dependency actually lives, asked of NODE rather than guessed from a path.
|
|
109
|
+
*
|
|
110
|
+
* 0.19.2 shipped a false FAIL to every third-party consumer: it built the directory by joining
|
|
111
|
+
* `packageRoot/node_modules/<name>`, and npm HOISTS dependencies to the project root, so for an
|
|
112
|
+
* install at `<proj>/node_modules/@holmes-lab/holmes-kit` the joined path never existed — while the
|
|
113
|
+
* binding sat at `<proj>/node_modules/better-sqlite3` and loaded fine. Hoisting is npm's default,
|
|
114
|
+
* which made the false verdict the default path of a normal install.
|
|
115
|
+
*
|
|
116
|
+
* `package.json` is the resolution target because we want the DIRECTORY, not the entry point: a
|
|
117
|
+
* package whose `exports` hides its main still resolves its own manifest by convention. A failure
|
|
118
|
+
* to resolve returns null, and the caller reports "absent" exactly as before — the repair changes
|
|
119
|
+
* how we look, never whether a real miss is reported.
|
|
120
|
+
*/
|
|
121
|
+
export declare function resolveDepDir(fromRoot: string, name: string): string | null;
|
|
122
|
+
export declare function gatherNativeEvidence(packageRoot: string): NativeEvidence;
|
|
85
123
|
/** The parent environment minus the variables that legitimately change a gate decision. Pure. */
|
|
86
124
|
export declare function probeEnv(parent: NodeJS.ProcessEnv): NodeJS.ProcessEnv;
|
|
87
125
|
/** Seam for the gate probe's child process. Default is a real `spawnSync`; tests inject a fake. */
|
|
@@ -94,6 +132,18 @@ export interface ProbeRunner {
|
|
|
94
132
|
stderr?: string;
|
|
95
133
|
};
|
|
96
134
|
}
|
|
135
|
+
/**
|
|
136
|
+
* Where a wiring's executable actually is, judged from the PROJECT rather than from the caller.
|
|
137
|
+
*
|
|
138
|
+
* Measured 2026-09-10: the same `.codex/config.toml` under the same `--target` read PASS from the
|
|
139
|
+
* repository root and FAIL from `/tmp`, because a relative arg is resolved by Node against
|
|
140
|
+
* `process.cwd()`. A relative arg is the CONVENTION here — `.mcp.json` has always carried
|
|
141
|
+
* `bin/holmes-mcp.js` — so the verdict was decided by where the person diagnosing stood.
|
|
142
|
+
*
|
|
143
|
+
* An absolute arg is returned untouched: agy's wiring is absolute and must not move. An empty arg
|
|
144
|
+
* stays empty — resolving it would conjure the target directory itself into a "file that exists".
|
|
145
|
+
*/
|
|
146
|
+
export declare function resolveWiringPath(target: string, arg: string): string;
|
|
97
147
|
export declare function runDoctor(packageRoot: string, target?: string, opts?: DoctorOptions, extraChecks?: Check[]): Promise<Check[]>;
|
|
98
148
|
/**
|
|
99
149
|
* Drive a real MCP stdio handshake: initialize -> initialized -> tools/list, with a timeout.
|
|
@@ -101,7 +151,7 @@ export declare function runDoctor(packageRoot: string, target?: string, opts?: D
|
|
|
101
151
|
* and closes stdin makes the server exit before answering (a false FAIL this check produced on its
|
|
102
152
|
* very first run against a healthy server).
|
|
103
153
|
*/
|
|
104
|
-
export declare function wiringSpawnCheck(command: string, args: string[], timeoutMs?: number, platform?: NodeJS.Platform): Promise<Check>;
|
|
154
|
+
export declare function wiringSpawnCheck(command: string, args: string[], timeoutMs?: number, platform?: NodeJS.Platform, cwd?: string): Promise<Check>;
|
|
105
155
|
/**
|
|
106
156
|
* Push-gate presence, diagnosed only where it applies: a repo that opted into governance (.ax)
|
|
107
157
|
* AND has git. Absence or a hook without our signature is a WARN carrying the install command —
|