Addresses codex review feedback (P2). `resolve(__dirname, '../..')` breaks
when `skills/understand/` is copied to a runtime skills directory whose
parent is not the plugin checkout — exactly the case SKILL.md Phase 0 warns
about and resolves via its multi-candidate $PLUGIN_ROOT search.
This script now prefers $PLUGIN_ROOT from the env (validated via
package.json presence) and falls back to the existing relative resolution.
SKILL.md Phase 0.5 passes the env var in the invocation.
Same latent pattern exists in scan-project / compute-batches / extract-import-map
/ extract-structure / build-fingerprints; hardening those is a separate
concern (no behaviour change today for installs that already work).
Refs #76
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the duplicated Node.js block in Phase 0.5 with a call into
`generateStarterIgnoreFile` via a thin wrapper script, mirroring the
scan-project.mjs pattern. Removes ~40 lines of duplicated logic; single
source of truth in @understand-anything/core.
Also tightens code review nits:
- Add 3 tests: stable language-group ordering, all-commented invariant
on empty dirs, suffix-glob rejects non-directory entries
- Clarify comments on EXACT_DIR_NAMES (ecosystem mix, not Python) and
SUFFIX_DIR_GLOBS (unanchored String.endsWith match)
- Type detectDirectories' readdirSync result explicitly (Dirent[]) to
pin the utf-8 encoding overload
Refs #76
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Detect C# project-suffix dirs (Foo.Tests/, Foo.UnitTests/) and PascalCase
test dirs (Tests/, UnitTests/, IntegrationTests/) via case-insensitive
match; group test-file suggestions by language (JS, C#, Java, Go).
Keeps all suggestions commented-out — same opt-in model as today. Updates
SKILL.md Phase 0.5 inline generator to stay in sync with the TS module.
Refs #76
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Mirror the English README's Kiro additions across all 7 translations
(es-ES, ja-JP, ko-KR, ru-RU, tr-TR, zh-CN, zh-TW): the one-line install
heading, supported <platform> values, a dedicated Kiro CLI / IDE
section, and the Platform Compatibility table row.
Addresses the maintainer (@Lum1104) change requests and the Codex
review on PR #364.
install.sh
- Build the Kiro agent's "resources" list dynamically by iterating over
understand-anything-plugin/agents/*.md instead of hard-coding each path
(review: "do not hard code the agents, instead use loop of the repo
dir"). Deterministic order via LC_ALL=C sort; pure bash, no jq
dependency. This also fixes pre-existing drift: the hard-coded list had
7 entries and was missing article-analyzer.md and
knowledge-graph-guide.md, while 9 agent files actually exist.
- Drop the unsupported "Kiro IDE auto-discovers .kiro-plugin/plugin.json"
install tip; keep the usage hint.
install.ps1 (Windows)
- Add the `kiro` platform entry (~/.kiro/skills, per-skill) so
`install.ps1 kiro` no longer fails with "Unknown platform: kiro".
- Generate ~/.kiro/agents/understand.json in Cmd-Install using the same
dynamic agent discovery; emit forward-slashed file:// URIs and write
UTF-8 without a BOM via [System.IO.File]::WriteAllText.
- Remove the agent JSON in Cmd-Uninstall.
README.md
- Point the Kiro install command at the canonical Egonex-AI repo instead
of Lum1104 (Codex P2: a lagging fork would error "Unknown platform:
kiro" before cloning the canonical repo).
- Replace the unverified ".kiro-plugin auto-discovery" claim with the
actual behavior (skills symlinked into ~/.kiro/skills, agent written to
~/.kiro/agents/understand.json).
Remove .kiro-plugin/plugin.json
- Kiro's documented discovery locations are ~/.kiro/agents, .kiro/skills
and .kiro/prompts only; .kiro-plugin/plugin.json is never read, so the
file and the claims that referenced it are removed (maintainer: "use
install.sh only is fine").
Verification
- install.sh: `bash -n` clean; executed the real agent-JSON generation
block -> valid JSON with 9 resources; `kiro-cli agent validate` exit 0.
- install.ps1: parsed with 0 syntax errors (PowerShell 7.6.2 AST parser);
ConvertTo-FileUri yields file:///C:/... for Windows paths; the real
Cmd-Install "kiro" path runs without error and produces JSON that
passes `kiro-cli agent validate` (exit 0).
Two gaps in the call-graph walker, both flagged by codex on #435:
1. `const Foo(...)` / `new Foo(...)` constructor calls were silently
dropped. The grammar emits these as `const_object_expression` /
`new_expression` containing `arguments` directly — they bypass the
`selector > argument_part` shape the walker relied on. Added a
dedicated branch that records the inner `type_identifier` as the
callee. Critical for Flutter widget trees where
`runApp(const MyApp())` would otherwise lose the MyApp construction
edge.
2. When a getter / setter / constructor / factory_constructor has a
body, its `method_signature` wraps `getter_signature` /
`setter_signature` / `constructor_signature` /
`factory_constructor_signature` instead of `function_signature`. The
walker only looked for `function_signature`, so `pendingName`
stayed null and the sibling `function_body` was walked with an
empty stack — calls inside ctor/factory/getter/setter bodies were
silently dropped even though those members were already extracted
as functions. Now dispatch across all five signature variants,
using `constructorName` for the (factory) constructor pair to
match what `collectClassBody` pushes.
Tests: 41 → 47 dart cases (+6); full core 733 → 739; no regressions.
CI runs `pnpm install` from the repo root using the root lockfile with
the default frozen-lockfile behaviour. The previous commits ran install
only inside understand-anything-plugin/ so the new
@understand-anything/tree-sitter-dart-wasm workspace dependency was
recorded in understand-anything-plugin/pnpm-lock.yaml but missed at
the repo root — CI would have stopped at install.
Flagged by @chatgpt-codex-connector on #435.
Incorporates stronger pieces from the prior Dart attempts (#348, #415)
that @Lum1104 called out:
- `extractParams` now walks `optional_formal_parameters` (covers both
optional positional `[...]` AND named `{...}` parameters — the Dart
grammar uses one wrapper for both).
- New `extractParamName` helper extracts the user-visible field name
from `this.field` and `super.field` initializer parameters by
unwrapping `constructor_param` / `super_formal_parameter`.
- `collectClassBody` now routes `getter_signature` and `setter_signature`
in both shapes:
- concrete: `method_signature > getter_signature` + sibling function_body
- abstract: `declaration > getter_signature`
Setters use the same path. The previous limitation assertion
(`methods).not.toContain("value")`) flipped to a positive
`.toContain("value")`.
- Added import/export edge-case tests: `dart:` SDK URIs, multi-import
declaration-order preservation, and `export ... show` clauses.
- Added a comma-list field test (`int a, b, c;`).
Underscore-prefix visibility carries through naturally to all new code
paths via the existing `isExported` gate inside `pushMethod`; explicit
test added for an underscore-prefixed getter.
Test counts: 28 → 41 dart cases; full core suite 720 → 733; no
regressions.
Per maintainer feedback on #436 — these are personal brainstorm/plan
artifacts produced via the superpowers skill flow, not repo
documentation. The BUILD.md provenance note in
packages/tree-sitter-dart-wasm/ stays since that's repo-level docs
about the vendored wasm.
Implements extractCallGraph with a sibling-aware walk that pairs each
function_signature with its subsequent function_body sibling (Dart's
AST differs from Kotlin's: signature and body are siblings, not
parent/child). Detects call sites via selector nodes containing
argument_part; uses startIndex for sibling lookup (web-tree-sitter
returns new wrapper objects per child() call, making === unreliable).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds enum_declaration handling to DartExtractor: enum constants are surfaced
as properties[] so the structural graph captures Color.red / Color.green etc.
Implements Task 9 of the Dart language support plan (TDD, 16/16 dart tests
pass, full suite 708/708).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add mixin_declaration handling to extractStructure, folding mixins into
classes[] (same convention as class_definition). The `on` constraint
sibling is intentionally ignored for graph purposes.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add constructorName() helper and extend collectClassBody() to surface
unnamed constructors as "ClassName", named constructors as "Class.named",
and factory named constructors as "Class.named" in methods[]/functions[].
Probe confirmed plan's AST shapes match exactly; extractReturnType returns
undefined for all constructor forms (factory keyword is an unnamed node).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add TDD tests and implement extractTopLevelFunction with helpers for
extracting function name, params, and return type (including generics
where the grammar emits type_identifier + type_arguments as siblings).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Empty extractor that satisfies the LanguageExtractor interface so the
plugin pipeline can load it. Real extraction logic lands in subsequent
TDD commits.
Adds the Dart language config and wires it into builtinLanguageConfigs
so .dart files are recognized by the language registry. References the
vendored @understand-anything/tree-sitter-dart-wasm package for grammar
loading.
No extractor yet — structural extraction lands in the next commit.
The upstream tree-sitter-dart@1.0.0 ships a pre-`dylink.0` wasm that
fails to load in web-tree-sitter@0.26.x. The grammar source itself is
sound — rebuilding with the current tree-sitter-cli + wasi-sdk produces
a working dylink.0 wasm. Vendor that artifact as a workspace-internal
package so @understand-anything/core can depend on it via workspace:*.
BUILD.md documents the provenance and rebuild instructions.
Thirteen-task TDD plan walking from vendoring the workspace wasm package
through scaffolding the extractor and adding extraction logic in
test-first slices: functions, classes, constructors, mixins, extensions,
enums, imports, exports, visibility, and call graph.
Every code block reflects AST shapes confirmed via a live probe against
a freshly-built tree-sitter-dart wasm in the project's own
web-tree-sitter at 0.26.x. No placeholder code, no "fill in later"
references.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Live verification during planning surfaced two facts that change the
shipping strategy:
1. tree-sitter-dart@1.0.0's prebuilt wasm uses the pre-`dylink.0` format
and fails to load in web-tree-sitter@0.26.x (the version this project
uses). Verified by directly loading the upstream wasm and catching
the failure in getDylinkMetadata.
2. The grammar source itself is sound — rebuilding with the current
tree-sitter-cli@0.26.x + wasi-sdk-29 toolchain produces a working
dylink.0-format wasm that parses every construct the extractor needs.
Revised packaging: ship the freshly-built wasm as a workspace-internal
package (@understand-anything/tree-sitter-dart-wasm) rather than
depending on the broken upstream npm artifact. No loader changes
required; existing TreeSitterPlugin resolves it the same way it
resolves other tree-sitter packages.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds the brainstormed design for landing deep Dart support at parity with
the recent Kotlin add (PR #347): LanguageConfig + tree-sitter WASM grammar
(tree-sitter-dart@1.0.0, verified ships a prebuilt .wasm in its tarball) +
DartExtractor + ~22 vitest cases. Six file changes, no edits to shared
schemas/registries.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Phase 7's `rm -rf` of the just-created `intermediate/` and `tmp/` dirs
trips destructive-action gates on hardened hosts (e.g. freshness-window
checks that flag deleting paths created moments earlier). Move them into
a timestamped `.trash-<epoch>/` instead; Phase 0 reclaims the space once
the trash is older than 7 days, well past any freshness window. Behavior
on normal hosts is unchanged — disk usage is identical after the next
run's purge.
Closes#301
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Addresses the regression flagged by ZebangCheng on #346: under the
parallelised `buildResolutionContext`, `loadTsConfigs` /
`loadGoModules` / `loadPhpAutoloads` ran concurrently but each wrote
warnings to stderr inline as it iterated read results, so a fixture
with both a malformed `tsconfig.json` and a malformed `composer.json`
could emit `composer, tsconfig` instead of the pre-PR `tsconfig,
composer` depending on I/O timing.
Each loader now buffers its warnings into a returned array and the
caller drains them in canonical order (tsconfig → go → php) after
`Promise.all`, restoring byte-identical stderr output. Added a
regression test that fixtures both malformed configs and asserts the
tsconfig warning precedes the composer warning in stderr.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
When /understand runs with no --language flag and no stored outputLanguage,
step 3.6 now infers the conversation language and — only when it is non-English
— confirms once before generating, then persists the choice to config.json.
English conversations keep the exact same silent `en` path; --language flag and
stored config still take priority. README documents the behavior; version
bumped 2.7.5 -> 2.7.6 across all five manifests (user-visible behavior change).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Add kiro to install.sh platforms_table (per-skill symlinks to ~/.kiro/skills/)
- Generate ~/.kiro/agents/understand.json on install for full pipeline support
- Clean up agent JSON on uninstall
- Add .kiro-plugin/plugin.json for IDE auto-discovery
- Update README with Kiro in platform table, one-line install list, and dedicated section
Resolve conflict in tests/skill/understand/test_extract_import_map.test.mjs
by keeping both new test groups — they cover independent fixes that should
coexist:
- upstream #214: tsconfig path-alias targets with leading "./"
- this PR #294: NodeNext .js → .ts rewrite for ESM TypeScript imports
The extract-import-map.mjs script auto-merged cleanly; both fixes are
already present in the merged source.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Fixes the silent near-edgeless-graph regression on any modern ESM
TypeScript project. Reported in #294 with full repro + root-cause
analysis.
### Why this matters
Under `moduleResolution: NodeNext` (or `Node16` / `Bundler` with
explicit extensions — the default for new TS-ESM projects since 2023),
TypeScript does NOT rewrite import specifiers during compilation:
// src/index.ts — real, idiomatic NodeNext source
import { x } from './config.js'; // on disk: config.ts
Before this fix, `probeWithExtensions` only tried APPENDING extensions
to the import specifier:
'./config.js' → not in fileSet
'./config.js.ts', './config.js.tsx', './config.js.js', ... → all miss
→ returns null → edge dropped at merge as dangling
Net result on the reporter's repro: a knowledge graph with hundreds of
file nodes and almost no `imports` edges between them — silently
removing exactly the dependency structure the graph is meant to show.
### Fix
New `NODENEXT_REWRITES` table maps each compiled-output extension to
the TypeScript source extensions that could have produced it:
.js → [.ts, .tsx, .js, .jsx]
.jsx → [.tsx, .jsx]
.mjs → [.mts, .mjs, .ts]
.cjs → [.cts, .cjs, .ts]
`probeWithExtensions` now applies the rewrite when the import already
ends with one of these extensions and no such file exists on disk. The
rewrite runs BEFORE the legacy append-extensions loop — otherwise
`./foo.js` would generate the nonsense candidate `foo.js.ts` and the
append loop would never reach the actual `foo.ts`.
### Disambiguation
If both `config.ts` and `config.js` exist on disk (rare, but possible
during a partial migration), `import './config.js'` still resolves to
the .js — that's an exact-disk match and what NodeNext compilation
actually does. The rewrite only kicks in when the .js doesn't exist.
### Tests
6 new tests in `test_extract_import_map.test.mjs`:
- The main #294 case (`.js → .ts`)
- `.jsx → .tsx` and `.mjs → .mts` rewrites
- Disambiguation when both `.ts` and `.js` exist on disk
- Pure-JS projects still work (real `.js → .js` imports)
- Historical no-extension probes unaffected
- Missing files still return null (rewrite can't invent targets)
Total: 202 tests passing (was 196).
Closes#294
Closes a cluster of community-profile gaps (#248, #249, #251, #252) in one
PR rather than four micro-PRs that all touch the same surface area.
### Templates (#251, #252)
- .github/ISSUE_TEMPLATE/bug_report.yml — required fields for repro
(plugin version, platform, OS, project language, file count); the four
pieces of context that are missing from ~every current bug report.
- .github/ISSUE_TEMPLATE/feature_request.yml — leads with the *problem*
rather than the proposed solution, which keeps maintainer review focused
on whether to solve, not just how.
- .github/ISSUE_TEMPLATE/question.yml — separate from bug to keep the
bug queue triagable.
- .github/ISSUE_TEMPLATE/config.yml — disables blank issues and routes
general discussion to README + Discussions.
- .github/PULL_REQUEST_TEMPLATE.md — includes the version-bump checklist
that CLAUDE.md says must stay in sync across 5 manifests; otherwise
every contributor learns this rule by getting their PR bounced.
### Community files
- CODE_OF_CONDUCT.md — short, project-specific document that names the
expectations and reporting path. Not a verbatim Contributor Covenant
to keep it readable.
- SECURITY.md — describes the project's local-only threat model
explicitly so reporters know what's in / out of scope before they
spend time on a writeup. Points at GitHub private vulnerability
reporting as the primary channel.
### CI (#249)
- ci.yml now also runs on pushes to main, not only PRs. Without this,
a direct push to main (which happens when maintainers merge a PR
branch locally) doesn't trigger CI, so a regression can land green-
looking and stay broken for days.
- Added a concurrency group that cancels stale runs for the same ref.
Saves runner minutes and keeps the per-ref status meaningful.
- Used `github.ref` (a controlled value), not user-controlled input,
so no script-injection surface.
### package.json (#248)
- Added description, license, repository, bugs, homepage, keywords —
the standard set for npm package discoverability and so GitHub's
community-profile check shows the project at 100%.
Lum1104
thejeshandClaude Opus 4.7
smjeong84
chienvon
chengyongru
ZebangChengandGitHub
Tirth KananiandClaude Opus 4.7
Bozheng LongandClaude Opus 4.8