CalvinBackup/claude-howto
1
0
Fork 0
mirror of https://github.com/luongnv89/claude-howto.git synced 2026-09-22 01:10:42 +02:00
Code Issues Packages Projects Releases Wiki Activity
248 Commits 1 Branch 10 Tags
8aeb5a72dc2fc105141949fcfc84c16ef35cfa47
Commit Graph
10 Commits
Author SHA1 Message Date
Luong NGUYEN 8aeb5a72dc docs: sync to Claude Code v2.1.278 (#183) 
* docs: sync to Claude Code v2.1.278

Type: docs

Audit of the tutorial against Claude Code 2.1.278 (repo baseline was
2.1.263). Fixes claims the repo asserts that are now false; new-feature
coverage is English-only.

P0 - the repo stated the opposite of current behavior:
- AGENTS.md: the claude-md skill said Claude Code "does not read
  AGENTS.md directly" in all 5 languages. It has since v2.1.277, when no
  CLAUDE.md exists at or above the working directory.
- /output-style: listed as "Removed in v2.1.91" in 5 languages and in
  09-advanced-features. Re-added in v2.1.269; row moved out of the
  Deprecated Commands table.
- Permission rules: setup-auto-mode-permissions.py seeded Write(*),
  Glob(*) and NotebookEdit(*), which Claude Code accepts but never
  consults and warns about at startup. Removed; same fix in the
  config-examples.json profile and the 06/09 category tables.
- taskOutputMaxChars: documented as live. The TaskOutput tool was
  removed in v2.1.278 and the key is inert.

P1:
- Dynamic workflow "medium" guideline is fewer than 10 agents, not 15;
  the default is small on Pro plans since v2.1.271.
- Todo-tool availability restated as the official allowlist (Claude 3.x,
  Opus 4-4.7, Sonnet 4-4.6, Haiku 4.5) instead of a denylist; 23 sites
  across 5 languages.
- Memory hierarchy now notes the AGENTS.md fallback.
- Auto-mode classifier cost qualified: not charged on Enterprise plans
  and Claude API accounts since v2.1.278.

P3 (English only): full AGENTS.md section in 02-memory; omitClaudeMd
subagent field; plugin eval / --json / --accept-command / --marketplace;
maxEffortLevel, bashEditDiffEnabled, syncClaudeAi*, 4 env vars,
--system-prompt-snapshot, --plugin-dir folders; send-now key and Monitor
deadlines; checkpoints subagent-edit limitation.

Learning skills audited clean: 100 quiz questions and 100 Review
pointers all valid, 20 self-assessment options correctly mapped.

Cross-file anchors: 15 broken before, 15 after, 0 introduced.

* docs: address review — stale version stamps, /output-style re-add, permission tables
 2026-09-20 01:15:34 +02:00
Luong NGUYEN 1e16914a8d docs: sync to Claude Code v2.1.257 (#180) 
* docs: sync to Claude Code v2.1.257

Audit against Claude Code v2.1.257 (repo baseline was v2.1.245, 12 releases
behind). Facts verified against the official docs and cross-checked against
the locally installed 2.1.257 binary.

Critical (P0):
- lesson-quiz L06 Q9 asked how many hook events exist, offering 16/25/31/40.
  With the real answer now 33, no option was correct. Option C 31 -> 33;
  answer letter unchanged.
- 03-skills taught SKILL.md `name`/`description` as required. Official docs
  now state all frontmatter fields are optional and `name` defaults to the
  directory name. Corrected in all five languages.

Hook events 31 -> 33 (PreModelSwitch/PostModelSwitch, v2.1.251) across 30
occurrences in 22 files, including three that read 25. Delta proved by
set-diff against the official event table: no phantom events. Both events
added to the 06-hooks tables with matcher, blocking behavior and 30s timeout,
plus detailed per-event sections.

CLI corrections verified against the binary:
- `claude plugin tag` takes a path, not a version
- `claude plugin init` / `validate` require their arguments
- `plugin init` scaffolds at ~/.claude/skills/<name>/, not project-local
- `--channels` entries must be tagged plugin:<name>@<marketplace>
- `--remote` is a deprecated alias for `--cloud`
- no `claude remote-control` subcommand exists, only the --remote-control flag

Also: CATALOG summary-table totals recomputed in all five languages (column 4
is a derived total, so the count change invalidated it); stale Opus 4.6/4.7
rosters aligned to canonical; self-assessment Advanced Features rescored 0-3
to stop a real gap being graded as mastery; 9 dangling quiz Review pointers
repaired; new features documented (--restricted, Fable 5.1,
experimental.cacheTtl, CLAUDE_CODE_SUBAGENT_MODEL_FORCE).

Deliberately out of scope: pre-existing translation staleness whose fix is
translation authorship rather than correction (ja/uk April-2026 root-doc
bodies, zh stubs, vi/README stub, translated CATALOG "New Features").

Claude-Session: https://claude.ai/code/session_01128xfEd8W7c2cMRw6N7uiF

* fix(hooks): complete the 31 → 33 hook-event propagation across all languages

The v2.1.257 sync bumped the hook-event count in the lesson and catalog
summaries but left 14 files still asserting 31, four of which the same
commit restamped 2.1.257. It also bumped the prose count in
claude_concepts_guide.md without adding the two rows, making that file
contradict itself where it had been internally consistent before.

Adds PreModelSwitch and PostModelSwitch (v2.1.251) everywhere an
exhaustive enumeration exists, and corrects the count wherever a list is
deliberately partial:

- README.md / INDEX.md in en, ja, zh, uk and vi/INDEX.md — count 31 → 33
  and both events added to the lifecycle group
- resources.md in all five languages — count only; these lists end in
  "and more" and are partial by design
- claude_concepts_guide.md — the two missing table rows

Also brings every hook-event table up to the 33 its own summary row
claims. The English CATALOG.md listed 29; the translated catalogs listed
25 and the ja/uk concept guides 27 and 28. Completing them required
translating six events that had never reached the mirrors: Setup,
UserPromptExpansion, PermissionDenied, PostToolBatch, MessageDisplay and
DirectoryAdded.

Summary-table arithmetic is untouched — those columns count example
scripts, not events.

Verified: no stale count survives outside CHANGELOG history, every
exhaustive enumeration lists exactly 33, and pre-commit passes 23/23.

Claude-Session: https://claude.ai/code/session_01UXmqJZNk1BR4uerJA85NiW
 2026-09-02 10:35:08 +02:00
Luong NGUYEN 28149a85fc docs: sync to Claude Code v2.1.245 (#177) 
* docs: sync to Claude Code v2.1.245

Sync the tutorial, its four translations, and both learner-facing skills
against Claude Code v2.1.245 (was anchored at v2.1.235).

The graded material turned out to be staler than the tutorials it grades.
The lesson READMEs already said "31 hook events", "six rewind options" and
"six permission modes"; the quiz and self-assessment still said 25, 5 and 5,
so learners were scored against a wrong answer and then sent to a lesson
that contradicted it. Seven such claims are fixed here.

Two long-standing doc defects were also verified against official docs:

- Hook default timeout was documented as 60 seconds. It is 600 for
  command/http/mcp_tool (30 for prompt, 60 for agent), with per-event
  overrides. Wrong in five places.
- Skill precedence was documented as enterprise > project > personal. The
  official order is enterprise > personal > project. question-bank.md
  Lesson 03 Q9 already graded the correct order, so the lesson had been
  contradicting its own quiz.

Fixed
- Hook default timeout 60s -> 600s, with per-event overrides
- Skill precedence: project and personal were swapped
- Memory "overrides root CLAUDE.md" -> concatenation model (last holdouts
  after the quiz was aligned in #164)
- permissionMode frontmatter omitted `auto` (English + vi/ja/uk); the quiz
  was right and the docs were wrong
- Built-in subagent `Bash` no longer exists upstream -> renamed to `claude`
- `/undo` is absent from the official commands reference -> marked deprecated
- `backgroundTasks` settings block was fabricated -> replaced with
  CLAUDE_CODE_DISABLE_BACKGROUND_TASKS and the concurrency env var
- Plugin manifest field is `lspServers`, not inline `lsp`
- vi/05-mcp documented `claude mcp add --transport ws`, which upstream
  explicitly does not support
- Self-assessment could never reach its own maximum score: rounds cap at
  4+4+4+3+4 = 19, not 20
- 13 dangling quiz `Review:` pointers, two of which named another lesson
- Skill description cap: 1024 chars -> 1,536 combined with `when_to_use`

Added
- WebSocket (`ws`) MCP transport, undocumented until now
- Concise output style (v2.1.237), a fifth built-in
- modelPicker, promptCacheTtl, subagentPromptCacheTtl, modelPricing,
  keybindingFlavor settings
- ANTHROPIC_DEFAULT_MODEL env var
- Hook `async`, `asyncRewake`, `shell`, `statusMessage`, `retry` keys;
  `agent` type marked experimental
- /design canvas, Remote Control out of research preview, notify_when_idle,
  plugin manifest fields (workflows, channels, dependencies, outputStyles)

Deferred deliberately: the reserved `synced` skills folder, and a bulk
footer stamp across the translations. zh/ carries none of the P0 claims
(it is abridged upstream of this run), so it has no P0 sub-tasks rather
than fabricated English.

Mermaid validation did not run in this environment (mmdc absent); CI covers it.

* fix(docs): propagate v2.1.245 corrections to CATALOG, translations, and stale enumerations

* fix(docs): propagate built-in subagent rename to uk/vi/zh CATALOG and zh lesson
 2026-08-26 07:20:04 +02:00
Luong NGUYEN deb7e5c710 docs: sync to Claude Code v2.1.233 (#168) 
* docs: sync to Claude Code v2.1.233

Bring the tutorial from the v2.1.220 baseline up to v2.1.233 (13 releases)
per update-plan-2026-08-15.md.

Corrections:
- Remove /ultraplan (removed in v2.1.222); point at plan mode instead
- Drop the 200-subagent-per-session cap and CLAUDE_CODE_MAX_SUBAGENTS_PER_SESSION
  (cap removed in v2.1.224)
- Note TaskCompleted/TaskCreated only fire when todo tools are enabled

Updates:
- /review documented as an alias of /code-review (v2.1.223)
- fork subagents default to interactive (v2.1.232)
- ultrareview v2.1.227 floor, --post/--no-post
- GitLab MR support for -w/--worktree and Agent View

New sections:
- Cross-session messaging (ListAgents, SendMessage, crossSessionInbound)
- Sandbox credential masking (v2.1.221, v2.1.224)
- Trust and permission scoping
- archive and command plugin sources, conditional plugin activation
- claude self-hosted-runner, 7 new env vars

Translations (vi/ja/zh/uk) mirrored where the corresponding pages exist.
Every modified file carries the 2.1.233 metadata footer; files already at
2.1.220 with no content change were left alone.

* docs: address review findings from the v2.1.233 sync

- Mirror the /review alias correction (v2.1.223) into vi/ja/uk/zh
  01-slash-commands: add the live alias row, drop the stale row from the
  deprecated table, and rewrite the history bullet. All four still called
  /review deprecated in favour of a code-review plugin while carrying a
  2.1.233 stamp.
- Rewrite the 10-cli CLAUDE_CODE_FORK_SUBAGENT row to agree with
  04-subagents: fork mode is on by default in interactive sessions on every
  build since v2.1.232; the variable enables it where it is off by default.
- Remove a stray blank line that split the 10-cli environment-variable
  table, leaving the seven new v2.1.221-v2.1.233 rows without a delimiter
  row and rendering them as literal pipe text.
- Add the todo-tools caveat to the TaskCompleted row alongside TaskCreated
  in claude_concepts_guide, uk/ja concepts guides, and the uk/vi/zh/ja
  CATALOGs.
- Fix the ja 07-plugins and 09-advanced-features footers, which paired a
  2.1.233 version stamp with an April 24 last-updated date.

* docs(subagents): correct the stale fork-subagent claim in ja pages (#171)

ja/10-cli and ja/04-subagents both stated that CLAUDE_CODE_FORK_SUBAGENT
only enables forked subagents on external builds (Bedrock, Vertex,
Foundry) with no effect on the Anthropic API. That was true at v2.1.117
but has been wrong since v2.1.232: fork mode is on by default in
interactive sessions on every build, first-party or not, and the variable
now turns it on where it is off by default (claude -p, the Agent SDK, or
Claude Code older than v2.1.232).

Both pages remain stamped 2.1.119 on purpose. Only the individual claim
was verified and corrected; bumping the page stamp would assert a
whole-page verification that has not been done. Do not bump these stamps
without a full resync of the pages.
 2026-08-17 12:44:44 +07:00
Luong NGUYEN 4f3fa85d7e docs: follow-ups from the v2.1.220 accuracy review (#161) 
* fix(ci): move the EPUB build out of pre-commit into CI

The build-epub hooks called scripts/build_epub.py, which raises when the
mmdc binary is missing, while check_mermaid.py skips with a warning in the
same situation. On arm64 — where @mermaid-js/mermaid-cli has no working
build — that made 'pre-commit run --all-files' impossible to satisfy for a
docs-only change without --no-verify.

Remove the build-epub, vietnamese-build-epub and japanese-build-epub hooks
and widen the CI build-epub job to a language matrix so PR-time coverage is
not reduced: it previously built en only, and now builds en, vi, zh and ja.
zh was covered by neither the hooks nor this job, so a broken zh diagram
could only surface at release time.

Closes #156

* docs: replace the stale Kroki narrative with local mmdc rendering

Commit e76bbe4 swapped the Kroki HTTP dependency for local mmdc rendering
but left the docs describing the old design. The claim appeared in seven
live files, not the two originally reported: CLAUDE.md, scripts/README.md
and their ja/uk/vi copies.

Also corrects what the same staleness dragged along:
- scripts/README.md documented --timeout and --max-concurrent, which no
  longer exist, and omitted --mmdc-path, --lang and --puppeteer-config
- 'Async concurrent fetching' described a render_all() that is now a plain
  sequential loop with a dedupe cache
- the network-error/rate-limiting troubleshooting entries are replaced with
  the failures that actually occur now (missing mmdc, Chromium sandbox)

CHANGELOG entries recording the Kroki-to-mmdc switch are left alone; they
describe the past accurately.

Closes #157

* fix(scripts): make the ruff config match files that exist

Ruff resolves include/per-file-ignores patterns relative to the directory
holding pyproject.toml. Since the config lives in scripts/, the pattern
"scripts/**/*.py" meant scripts/scripts/**/*.py, which matches nothing —
'ruff check scripts/' printed 'warning: No Python files found under the
given path(s)' and then 'All checks passed!'. per-file-ignores had the same
mistake. Both are now relative to scripts/, so ruff sees all 14 files.

This silently disabled the pre-commit ruff hooks too, not just the
documented command, and two test files had drifted out of format as a
result. Reformatting them exposed a second problem: the pre-commit hook
pinned ruff v0.8.2 while the CI lint job installs unpinned latest and this
venv has 0.15.10. The versions disagree on formatting, so each reverted the
other's output, and CI would have started failing on a file the local hook
kept rewriting. Bump the hook pin and the requirements-dev floor to 0.15.10
so all three agree.

Verified: 'pre-commit run --all-files' is stable across consecutive runs,
'ruff format --check scripts/' and 'ruff check scripts/' both pass, 92
tests pass.

Closes #160

* docs: point localized module links at their own trees

Files at the root of a language directory used ../NN-module/, which from
uk/CATALOG.md resolves to the English 05-mcp/ at the repo root rather than
uk/05-mcp/. Readers following a count stated about the localized tree
landed in the English one. The link checker never caught it because the
English target does exist.

Files one level deeper (uk/04-subagents/README.md and friends) were already
correct — ../06-hooks/ from there resolves to uk/06-hooks/ — so this is
scoped to the 7 depth-1 files that actually escape: CATALOG.md in uk, vi,
ja and zh, LEARNING-ROADMAP.md in uk and vi, and ja/claude_concepts_guide.md.
119 links in total.

The ../NN- occurrences left in STYLE_GUIDE.md and TRANSLATION_NOTES.md are
inside fenced examples showing what a module page should contain, where the
../ form is the correct convention. All 119 rewritten targets were verified
to exist.

Closes #158

* docs(hooks): stop conflating hook types with hook event categories

'**Hook Types** (5 types, 31 events)' sat above four bullets, so the count
read as wrong. It was not — the two numbers describe different axes that
had been merged into one label.

06-hooks/README.md:126 documents five hook *types*: command, http, prompt,
mcp_tool and agent. Those describe how a hook runs. The four bullets are
event *categories* — Tool, Session, Task, Lifecycle — holding 31 events
(6+7+6+12), and describe when it runs.

Split the label so each number belongs to the axis it counts, in all nine
affected files: README.md and INDEX.md plus the ja, uk, vi and zh copies.
QUICK_REFERENCE.md and LEARNING-ROADMAP.md already stated '5 types' with
the handler names attached and needed no change.

Closes #159

* chore(scripts): drop the unused httpx dependency

httpx was the HTTP client for Kroki rendering. Nothing has imported it
since e76bbe4 moved diagram rendering to a local mmdc subprocess —
check_links.py, the only other network caller, uses stdlib urllib.request.
The PEP 723 blocks in build_epub.py and build_website.py had already
dropped it; only the manifests and docs still declared it.

Removes it from requirements.txt, pyproject.toml dependencies, the
dependency tables and uv --with lines in scripts/README.md and its ja/uk
copies, and the .cspell.json word list. Also drops the B113 bandit
suppression, which existed solely for an httpx timeout false positive —
bandit reports no issues without it.

Also corrects two Requirements lines missed in the previous pass: the ja
and uk script READMEs still listed an internet connection rather than mmdc.

Note: tenacity is now dead for the same reason and is left in place.

* chore(scripts): drop the unused tenacity dependency

Nothing imports tenacity — the retry logic it was added for went away
with the Kroki HTTP fetching in e76bbe4. It was still declared in
requirements.txt, pyproject.toml, both `uv run --with` lines and all
three dependency tables.

* docs(i18n): sync the localized pre-commit check lists

The ja/uk/vi CLAUDE.md files still listed build-epub as pre-commit
check #5, stale since 6da8184 moved the EPUB build to CI. They also
omitted markdown-rendering, and the ja/uk stated counts disagreed with
their own lists. Now matches the English CLAUDE.md.

* fix(scripts): silence PLR0917 now stable in ruff 0.16

CI installs the latest ruff (uv pip install ruff, unpinned), and 0.16.1
promoted too-many-positional-arguments from preview to stable, hard-failing
the Code Quality check on build_epub.py's long-signature draw/helper
functions. These sit in the same family as PLR0913, which is already
ignored. Add PLR0917 so the local 0.15.10 pin and unpinned CI lint agree
again.
 2026-08-06 07:58:46 +07:00
Luong NGUYEN b9a973bf32 docs: accuracy pass against Claude Code v2.1.220 (#155) 
Internal accuracy pass against v2.1.220 — no missing upstream features, but broken example code, disagreeing counts, and metadata drift.

Functional fixes: pre-commit.sh now exits 2 so it actually blocks; dependency-check.sh reads file_path from stdin JSON instead of $1; database-mcp.json uses ${DATABASE_URL}; broken fences repaired; three command templates had invalid skill names.

Factual corrections: /fork and /subtask unswapped and /subtask added; /fewer-permission-prompts; permissions.defaultMode; dontAsk/auto unreversed; 31 hook events verified name-by-name; subagent depth 3; skill precedence enterprise > project > personal; /output-style removed not deprecated; permissionDecision gained defer.

Follow-up review fixed defects the pass left behind: zh/vi headers claiming 31 events above 25-name lists, a surviving hardcoded DB credential in the MCP README examples, an unbalanced fence swallowing a metadata footer, and non-canonical tool names. All four translated CATALOG summary tables were recounted so their arithmetic holds.

Full detail in CHANGELOG.md under v2.1.220-r2.
 2026-08-04 15:41:12 +07:00
Luong NGUYEN c7261394b0 fix(i18n): re-point translations to code-review-specialist rename (#128) 
* fix(i18n): re-point translations to code-review-specialist rename

The English code-review skill was renamed to code-review-specialist in #127
to avoid shadowing the new built-in /code-review command. Mirror that rename
across the uk, vi, ja, zh locales: rename each 03-skills/code-review/ directory
to code-review-specialist/, update the i18n-source markers in the moved ja
files, and fix path references in INDEX.md, QUICK_REFERENCE.md, and the vi
TRANSLATION_QUEUE.md.

* chore: ignore .asm-improver and .gitissue tooling state

* fix(i18n): re-point remaining code-review install paths to code-review-specialist
 2026-05-26 10:29:51 +02:00
Luong NGUYEN cf92e8e2c1 chore(release): v2.1.112 
Sync all English tutorials with Claude Code v2.1.112 and the new Opus 4.7
model (claude-opus-4-7). Introduces the xhigh effort level (new default on
Opus 4.7), two new built-in slash commands (/ultrareview,
/less-permission-prompts), auto mode without --enable-auto-mode for Max
subscribers on Opus 4.7, PowerShell tool on Windows, Auto-match-terminal
theme, and plan files named after prompts. All 18 EN doc footers bumped to
Claude Code v2.1.112.

Version callouts in README.md, zh/README.md, and uk/README.md updated to
v2.1.112. Changelog entry prepended.

Repo version now tracks Claude Code version (2.3.0 → 2.1.112 is intentional
per new convention).
 2026-04-16 23:13:34 +02:00
Evgenij I f6d73e2659 fix(uk): replace stub README with full translation, fix broken anchors 
- Replace 3KB stub README with full 31KB translation matching original structure
- Fix apostrophe anchor mismatch in 04-subagents (U+02BC vs U+0027)
- Fix duplicate heading in 06-hooks (Prompt-хуки → Хуки на основі промптів)

Ref: luongnv89/claude-howto#63
 2026-04-10 08:54:06 +03:00
Evgenij I 06caeaa8d9 feat(i18n): add Ukrainian localization infrastructure 
- uk/README.md: Ukrainian landing page with module index
- uk/TRANSLATION_NOTES.md: glossary and style guide
- uk/TRANSLATION_QUEUE.md: progress tracker (0/67 files)
- Directory structure for all 10 modules

Ref: luongnv89/claude-howto#63
 2026-04-09 13:12:52 +03:00