The installer now finds module.yaml in both skills/ and src/ directories,
including one level deep in subfolders. Updates bmb module-definition to
skills/module.yaml to match its actual structure.
bmad-builder reverted its skills/ directory back to src/ for installer
compatibility (bmad-code-org/bmad-builder#40). Update the external
modules manifest to match.
* refactor(installer): restructure installer with clean separation of concerns
Move tools/cli/ to tools/installer/ with major structural cleanup:
- InstallPaths async factory for path resolution and directory creation
- Config value object (frozen) replaces mutable config bag
- ExistingInstall value object replaces stateful Detector class
- OfficialModules + CustomModules + ExternalModuleManager replace monolithic ModuleManager
- install() is prompt-free; all user interaction in ui.js
- Update state returned explicitly instead of mutating customConfig
- Delete dead code: dependency-resolver, _base-ide, IdeConfigManager,
platform-codes helpers, npx wrapper, xml-utils
- Flatten directory structure: custom/handler → custom-handler,
tools/cli/ → tools/installer/, lib/ directories removed
- Update all path references in package.json, tests, CI, and docs
* fix(installer): guard ExistingInstall.version and surface module.yaml errors
Guard ExistingInstall.version access with .installed check in
uninstall.js, ui.js, and installer.js to prevent throwing on
empty/partial _bmad dirs. Surface invalid module.yaml parse errors
as warnings instead of silently returning empty results.
* fix(bmad-init): correctly resolve output_folder paths outside project root
When output_folder was set to an absolute path (e.g. /Users/me/outputs),
the {project-root}/{value} result template stored it as
{project-root}//absolute/path. resolve_project_root_placeholder then did
a naive string replace, producing /project//absolute/path — a broken path
that workflows could not resolve.
For relative paths outside the root (e.g. ../../sibling), the same naive
replace left un-normalized paths like /project/../../sibling in the
resolved config, which some tools mishandled.
Fix resolve_project_root_placeholder to strip the {project-root} token,
detect whether the remainder is absolute (returning it directly) or
relative (joining with project root and normalizing via os.path.normpath).
Fix apply_result_template to skip the template entirely when raw_value is
already an absolute path, and to normalize the result for relative-but-
outside paths. This covers the bmad-init SKILL write path, which bakes
the resolved path directly into config.yaml.
Add 7 tests covering all three path cases (absolute, relative-with-
traversal, normal in-project) for both functions.
* Address review comments
---------
Co-authored-by: Akhilesh Tyagi <akhilesh.t@nextiva.com>
Co-authored-by: Brian <bmadcode@gmail.com>
Merge license, homepage, repository, keywords, and author from
plugin.json into marketplace.json and remove the redundant file.
The npx skills installer only reads marketplace.json for skill
discovery — plugin.json contributed no functional value.
Two bugs combined to produce an empty agent-manifest.csv:
1. collectAgents() only scanned {module}/agents/ directories, but agents
live at various paths (bmm/1-analysis/bmad-agent-analyst/,
cis/skills/bmad-cis-agent-*, etc.). Now walks the full module tree.
2. All 9 BMM agent manifests declared type: skill instead of type: agent.
The manifest generator requires type: agent to include a directory in
agent-manifest.csv. CIS, GDS, TEA, and WDS already had the correct type.
Changes:
- Fix 9 BMM bmad-skill-manifest.yaml files: type: skill → type: agent
- Replace collectAgents/getAgentsFromDir with full-tree recursive scan
- Module field from manifest file always takes precedence over directory
- Remove dead skillManifest load (legacy .md agent support removed)
- Add TODO in bmad-artifacts.js documenting legacy agent pipeline as dead code
- Add 10 regression tests covering BMM, CIS, and GDS directory layouts
Replace execSync with execFileSync in npx wrapper so that
argument values containing spaces (e.g. --user-name "CI Bot")
are passed as discrete array elements instead of being split
by the shell.
Fixes#2066
Signed-off-by: majiayu000 <1835304752@qq.com>
Co-authored-by: Alex Verkhovsky <alexey.verkhovsky@gmail.com>
The frontmatter `title` field is the single source of truth.
The duplicate `# {title}` H1 heading was redundant and has been removed.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Short-circuit evaluation in step-01: explicit argument → conversation
context → full artifact scan. Stops prompting as soon as intent is
unambiguous. All existing scan behaviors preserved in tier 3.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds a Self-Check subsection at the end of step-03 that forces the
implementing agent to verify all tasks are complete and mark checkboxes
before handing off to the review step.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs(zh-cn): refine established project guides
clarify the boundary between new-project and established-project usage so zh-cn readers can choose the right workflow path.
align project context terminology and command references with current conventions while keeping guidance concise and executable.
Feishu: https://www.feishu.cn/
Made-with: Cursor
* docs(zh-cn): fix project-context syntax and wording
我将 project-context 文档中的提示块语法统一回项目规范的 `:::...:::` 形式，避免与文档风格约定不一致。
我同时把“在不同用户故事（story）间决策不一致”调整为“之间决策不一致”，提升中文表达的自然度与可读性。
Feishu: <https://www.feishu.cn/>
Made-with: Cursor
---------
Co-authored-by: leon <leon.liang@hairobotics.com>
Co-authored-by: Alex Verkhovsky <alexey.verkhovsky@gmail.com>
Agent skills referenced with shortened names (bmad-pm, bmad-architect,
etc.) that don't match installed skill names. Fixed to use actual names
(bmad-agent-pm, bmad-agent-architect, etc.) across EN, ZH-CN, and FR.
Also fixed bmad-research to three specific research skills (EN, FR) and
bmad-product-brief to bmad-create-product-brief (ZH-CN).
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs(zh-cn): refine help and quick-fixes guides
improve zh-cn troubleshooting guidance so users can quickly choose the right support path and self-recover from common issues.
align bmad-help and quick-dev usage wording with current invocation conventions and remove glossary-style appendices to keep the pages action-oriented.
Feishu: N/A
Made-with: Cursor
* docs(zh-cn): fix tip admonition fence syntax
我把 get-answers-about-bmad 文档中的 `::::tip` 语法改为仓库统一使用的 `:::tip`。
此前四冒号写法与项目文档约定不一致，可能导致提示块渲染异常；现在与现有 Starlight 写法保持一致。
Feishu: <https://www.feishu.cn/>
Made-with: Cursor
---------
Co-authored-by: leon <leon.liang@hairobotics.com>
* fix(docs): correct Hasselhoff spelling, add locale-aware 404 redirect
Fix "Hasslehoff" → "Hasselhoff" typo in customize-bmad.md across all
three locales (en, zh-cn, fr).
Add client-side locale detection to 404.astro so GitHub Pages serves
the correct localized 404 page instead of always showing English.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(build): exclude translated locales from llms-full.txt
llms-full.txt was including zh-cn and fr docs, tripling the content
with duplicate information in different languages. Restrict to English
only — translations add no value for LLM context consumption.
Reduces output from ~393K to ~114K chars (~29k tokens).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* refactor(i18n): extract locale config to shared module
Move locale definitions from astro.config.mjs into a shared
website/src/lib/locales.mjs consumed by astro config, build-docs,
and 404.astro. Adding a new locale is now a single-file change.
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(quick-dev): use absolute paths in code -r invocations
Agent CWD may differ from the project root in worktree setups,
causing relative paths to silently fail. Resolve paths via
git rev-parse --show-toplevel before invoking code -r.
* fix(quick-dev): add CWD fallback when git rev-parse fails
Adds graceful fallback to current working directory when
git rev-parse --show-toplevel fails (VCS unavailable).
* fix(quick-dev): make file path references clickable
Spec-file links use paths relative to the spec file's
directory (clickable in VS Code). Terminal output paths
use CWD-relative format for terminal clickability.
* fix(quick-dev): add :line suffix to step-oneshot path example
Aligns the file path example in step-oneshot.md with the clickable
`:line` format already enforced in step-03-implement.md and
step-05-present.md.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* refactor(installer): discover skills by SKILL.md instead of manifest YAML
Switch skill discovery gate from requiring bmad-skill-manifest.yaml with
type: skill to detecting any directory with a valid SKILL.md (frontmatter
name + description, name matches directory name). Delete 34 stub manifests
that carried no data beyond type: skill. Agent manifests (9) are retained
for persona metadata consumed by agent-manifest.csv.
* refactor(installer): remove dead task/tool/workflow manifest code
The remove-skill-manifest-yaml branch deleted the scanners that
discover tasks, tools, and workflows but left behind the code that
writes their manifest CSVs. Remove collectTasks/Tools/Workflows,
writeTaskManifest/ToolManifest/WorkflowManifest, their helpers, and
the now-unreachable getPreservedCsvRows/upgradeRowToSchema methods.
Update installer pre-registration and test assertions accordingly.
Switch skill discovery gate from requiring bmad-skill-manifest.yaml with
type: skill to detecting any directory with a valid SKILL.md (frontmatter
name + description, name matches directory name). Delete 34 stub manifests
that carried no data beyond type: skill. Agent manifests (9) are retained
for persona metadata consumed by agent-manifest.csv.
* refactor(installer): remove dead agent compilation pipeline
Delete 9 files (~2,600 lines) that compiled .agent.yaml to .md.
No .agent.yaml files exist in the source tree — agents now ship
as pre-built SKILL.md. Clean up all references in installer,
module manager, custom handler, base IDE, UI, and tests.
* refactor(custom-handler): remove dead install/copy/find methods
CustomHandler.install(), copyDirectory(), and findFilesRecursively()
are never called — custom modules are installed via moduleManager.install()
since Dec 2025. Also removes unused FileOps import and constructor.
Verified with before/after clean-installer comparison (codex + custom
modules with custom.yaml): output is identical.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(installer): remove dead compilation refs from docs and module manager
Address review findings from PR #2080 triage:
- Remove compile-agents from CLI action docs (en, fr, zh-cn)
- Remove dead vendorCrossModuleWorkflows() and .agent.yaml skip logic
- Clean stale compilation-era comments in manifest-generator
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* refactor(installer): remove legacy workflow, task, and agent IDE generators
All platforms now use skill_format exclusively. The old
WorkflowCommandGenerator, TaskToolCommandGenerator, and
AgentCommandGenerator code paths in _config-driven.js were
no-ops — collectSkills claims every directory before the
legacy collectors run, making their manifests empty.
Removed:
- workflow-command-generator.js (deleted)
- task-tool-command-generator.js (deleted)
- writeAgentArtifacts, writeWorkflowArtifacts, writeTaskToolArtifacts
- AgentCommandGenerator import from _config-driven.js
- Legacy artifact_types/agents/workflows/tasks result fields
Simplified installToTarget, installToMultipleTargets, printSummary,
and IDE manager detail builder to skills-only.
Updated test fixture to use SKILL.md format instead of old agent format.
* fix(installer): address PR review findings from #2078
- Fix temp dir leak in test fixture cleanup (use path.dirname)
- Fail loudly when skill_format missing instead of silent success
- Add workflow.md to test fixture for verbatim-copy coverage
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Inline workflow.md content directly into SKILL.md for: editorial-review-prose,
editorial-review-structure, help, index-docs, review-adversarial-general,
review-edge-case-hunter, and shard-doc. Deletes the now-redundant workflow.md
files. No behavioral change — same pattern as advanced-elicitation in PR #2076.
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Merge workflow.md content directly into SKILL.md and delete the
now-redundant workflow file. The frontmatter `agent_party` variable
moves into SKILL.md; all relative file references (`./methods.csv`)
remain valid.
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* fix(code-review): update sprint-status to done after review completes
The code-review workflow ended without updating sprint-status.yaml from
"review" to "done", leaving stories stuck in review status. The dev-story
workflow implies code-review handles this transition but it was dropped
during the v6.2.0 step-file architecture refactor.
- Add sprint_status path to workflow initialization
- Track story_key in step-01 when discovered from sprint status
- Add step-04 section 6 to update sprint-status.yaml and story file
- Add step-04 section 7 with next-step options
Closes#2043
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(code-review): address PR review findings — split gating, story_key guard, HALT
- Split section 6 guard: story file status gated on spec_file only,
sprint-status sync sub-gated on story_key separately
- Add conditional branch for manual choice in multi-story path so
story_key is cleared when user declines a story selection
- Add HALT directive after Next steps menu to prevent LLM runaway
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(code-review): restore actionable review output with interactive choices
The March 15 rewrite (PR #2007) removed the ability to auto-fix patches,
create action items in story files, and handle deferred/spec findings.
This restores interactive post-review actions:
- Deferred findings: auto-written to deferred-work.md and checked off in story
- Intent gap/bad spec: conversation with downgrade-to-patch, patch-spec,
reset-to-ready-for-dev, or dismiss options
- Patch findings: fix automatically, create action items, or show details
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* refactor(code-review): simplify triage to decision-needed/patch/defer/dismiss
Replace 5-bucket classification (intent_gap, bad_spec, patch, defer, reject)
with 4 pragmatic buckets. Findings always written to story file first.
Decision-needed findings gate patch handling — resolve ambiguity before fixing.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(code-review): address PR review findings in step-04-present
Replace undefined curly-brace placeholders with angle-bracket syntax,
add HALT guard before patch menu, guard spec_file references for
no-spec mode, and backtick category names for consistency.
* feat(code-review): add HALT guards, batch option, defer reason, final summary
Add strong HALT guards after decision-needed and patch menus to prevent
auto-progression. Add batch-apply option 0 for >3 patch findings. Prompt
for defer reason and append to story file and deferred-work.md. Show
boxed final summary with counts. Polish clean-review shortcut in triage.
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Brian
Alex Verkhovsky
Akhilesh Tyagi
github-actions[bot]
梁山河
lif
Murat K Ozcan
Emmanuel Atsé
dependabot[bot]