* fix(installer): preserve module-help.csv schema in merged bmad-help.csv (#2278)
The installer's mergeModuleHelpCatalogs was rewriting the merged catalog
under a different schema (module,phase,name,code,sequence,workflow-file,...)
than the documented source schema in every module's module-help.csv
(module,skill,display-name,menu-code,description,action,args,phase,...).
Worse, the parsing assumed the wrong source column order, so column data
was scrambled in the merged output. SKILL.md docs the source schema, so
the bmad-help skill was navigating a catalog whose actual columns no
longer matched its mental model.
Drop the transformation and the agent enrichment columns (which had no
consumers anywhere in the codebase). Emit rows verbatim in the source
schema, padding short rows and filling empty module fields. Sort by
module then phase, stable within phase to preserve authored order.
Closes#2278
* fix(catalog): normalize module-help.csv rows to documented 13-column schema
Many rows in core-skills/module-help.csv and bmm-skills/module-help.csv
were missing one column between description and phase, leaving them at
12 fields instead of 13. CSV consumers that read by header position
were silently mapping data into the wrong columns (description into
action, phase into args, required into before, etc).
Inserted an empty cell at column index 5 across all 31 affected rows
to restore alignment with the documented header
(module,skill,display-name,menu-code,description,action,args,phase,
after,before,required,output-location,outputs).
* fix(config): promote project_name to core, fixes#2279
project_name was a bmm-specific prompt despite being a universal
project-level concept used by every module — including core skills like
bmad-brainstorming, which loads from _bmad/core/config.yaml and was
silently broken because project_name lived under bmm. Users without bmm
installed could not run brainstorming at all.
Move:
- src/core-skills/module.yaml: declare project_name with prompt
"What is your project called?" and default {directory_name}, matching
what bmm previously had.
- src/bmm-skills/module.yaml: remove the bmm definition; add project_name
to the "Variables from Core Config inserted" header comment so
contributors can see what's inherited.
Migration for existing installs:
- tools/installer/modules/official-modules.js: after loadExistingConfig
reads each per-module config.yaml, hoist any keys that are now declared
in core but appear under non-core modules. Without this, the partition
logic in writeCentralConfig (which strips core keys from non-core
buckets) would silently drop the user's prior project_name on the next
quick-update. Generic — handles project_name today and any future
module→core promotions.
- The hoist preserves precedence: an existing core value beats a stale
module-side copy.
--yes seed:
- tools/installer/ui.js: add project_name to the hardcoded core seed
(using path.basename(directory) to match the {directory_name} default)
so non-interactive fresh installs populate it. Without this the seed
silently omits project_name and core skills fall back to literals.
Tests:
- test/test-installation-components.js Suite 43 (9 assertions) covers
the schema move, the loadExistingConfig hoist, and the precedence rule.
- Suite 35 fixture updated: project_name moved from bmm bucket to core,
with a stale bmm copy left in place to verify it gets stripped.
Verified manually:
- Fresh install -y: project_name lands in [core] of config.toml.
- Existing install with project_name in bmm/config.yaml: quick-update
hoists it to [core] and strips it from [modules.bmm].
* fix(installer): harden config-load against malformed config.yaml
Per augment review on #2348: loadExistingConfig stored any truthy
yaml.parse result (including scalars like '42'), which would later crash
_hoistCoreKeysFromLegacyModuleConfigs at \`key in cfg\` with
"Cannot use 'in' operator to search for ... in 42".
- loadExistingConfig: only keep parses that are plain objects (not
scalars or arrays). A corrupt config.yaml is now treated the same as
a parse error — skipped, not crashed-on.
- _hoistCoreKeysFromLegacyModuleConfigs: belt-and-suspenders type guards
on _existingConfig.core (in case it's populated by some other path)
and on each module cfg in the loop.
- Test Suite 43 adds 2 assertions covering a scalar core/config.yaml:
loadExistingConfig must not crash, and bmm.project_name must still
hoist into a clean core bucket.
step-07-validation template shipped with all 16 completeness checkboxes
pre-checked and Overall Status hard-coded to READY FOR IMPLEMENTATION,
defeating the gate. Reset checkboxes to unchecked, replace status with a
templated choice tied to the checklist and gap analysis, and instruct the
agent to only mark items the validation actually confirms.
Closes#2292
* fix(installer): require --tools for fresh --yes installs; remove --tools none (closes#2326)
Fresh non-interactive installs without --tools previously produced a
config-only install (~35 files vs ~1400 in the manifest) with no warning
and a "BMAD is ready to use" success card, leaving slash commands
unreachable. --tools none was an explicit opt-in for the same broken
state.
Now: fresh install + -y without --tools throws a helpful error pointing
at --list-tools. --tools none is rejected as an unknown ID. Empty and
typo'd tool IDs are also rejected. Existing-install paths (--action
update, quick-update, modify) are unchanged - they continue to reuse
previously-configured tools when --tools is omitted.
Adds --list-tools flag that prints all 42 supported tool IDs (id, name,
target_dir, preferred star) sourced from platform-codes.yaml.
English docs updated; localized docs (vi-vn, fr, cs, etc.) will sync via
the normal translation pass.
* fix(installer): address review for #2326 — single source of truth, drop dead code, add tests
- Refactor formatPlatformList to use IdeManager so --list-tools and --tools
validation see the same set of platforms. Eliminates the drift where suspended
platforms appeared in --list-tools but were rejected at validation.
- Drop unused getValidPlatformIds export.
- Flatten redundant block scope around the throw in the --yes-without-tools
branch (refactor leftover).
- Drop dead String() defensive cast (Commander always passes a string).
- Add Test Suite 42: 8 unit tests covering _parseToolsFlag empty/whitespace/
unknown/typo cases plus an integration check that --list-tools output and
--tools validation agree on the ID set.
* fix(installer): close --tools "" bypass and drop hardcoded tool count
- Replace truthy `if (options.tools)` guard with `!== undefined` in both
upgrade and fresh-install branches. Empty string now reaches
_parseToolsFlag and produces the specific "passed empty" error
instead of falling through to a generic message (fresh-install) or
being silently ignored (existing-install).
- Drop the hardcoded "42 supported tools" count from the prereqs in
install-bmad.md so the doc doesn't drift as platform-codes.yaml
changes.
Addresses augment / coderabbit review on #2346.
* fix: use full update path when --custom-source is passed with --yes
When --yes is used on an existing install, the installer auto-selects
quick-update. However, quick-update never re-clones custom module repos
— it only reads whatever is already in the cache. This means
--custom-source with a new version tag (e.g. @1.1.0) is silently
ignored and the previously cached version (e.g. 1.0.1) is reported as
"already up to date".
Default to the full update path when --custom-source is present, so the
custom repo gets re-cloned at the requested version. Also ensure all
installed modules are included in the selection when --yes is combined
with --custom-source, preventing previously installed modules from being
removed.
* fix: address review feedback on choices.find() and comment clarity
* style: prettier fix for empty-body methods in custom-module-manager
---------
Co-authored-by: Brian <bmadcode@gmail.com>
* fix(installer): preserve http protocol in custom module clone URLs
Previously, parseSource() hardcoded 'https://' when building cloneUrl,
forcing http:// Git URLs (e.g., internal LAN hosts) to upgrade to https.
This broke cloning for self-hosted Git servers that only serve over HTTP.
- Capture the protocol from the regex match instead of discarding it
- Update JSDoc and inline comments to document HTTP support
- Update install-custom-modules docs (EN, ZH, VN) to list HTTP URL type
Fixes the --custom-source flag for http:// addresses.
* docs(installer): update JSDoc to mention HTTP support in cloneRepo
Add HTTP to the cloneRepo method's JSDoc param description.
Also fixes minor spacing in empty arrow functions (formatting).
* docs(installer): fix JSDoc annotation for cloneRepo param
Correct @param backtick escaping in cloneRepo JSDoc.
Also documents HTTP as a supported protocol alongside HTTPS and SSH.
---------
Co-authored-by: 关惠民 <9155544@qq.com>
* fix(installer): route community installs through PluginResolver when marketplace.json ships
Community-catalog installs ignored .claude-plugin/marketplace.json, so modules
that nest module.yaml inside a setup skill's assets/ directory (e.g. Strategy 2
in PluginResolver) ended up half-installed: only module-help.csv and the
generated config.yaml landed in _bmad/<code>/, while the actual skill source
trees and module.yaml never got copied. The install would silently emit
"could not locate module.yaml" warnings and leave .agents/skills/ without
the module's skills.
The fix wires the existing PluginResolver onto the community path:
- CommunityModuleManager.cloneModule now detects marketplace.json after the
clone+ref-checkout completes and runs PluginResolver. The resolution is
stamped with channel/sha/registryApprovedTag/registryApprovedSha and cached
in _pluginResolutions, mirroring the existing _resolutions cache.
- OfficialModules.install consults the community plugin resolution and
delegates to installFromResolution (the same code path custom-source
installs already use). installFromResolution branches on communitySource
to write source: 'community' with the registry's approved tag/sha and
channel.
- resolveInstalledModuleYaml now searches the community-modules cache root
in addition to the external-modules cache, and the BMB setup-skill detector
walks src/skills/ and skills/ (not just the repo root) so collectAgents
FromModuleYaml and writeCentralConfig can find module.yaml in nested
marketplace-plugin layouts.
Backward compatibility: repos without marketplace.json (e.g. WDS, which
declares module_definition: src/module.yaml at the root) continue through
the legacy findModuleSource path with no behavior change. Verified against
the live zarlor/suno-band-manager community module and a 23-check fixture
suite covering Suno-shape, WDS-shape, and bare-repo layouts.
* fix(installer): harden community marketplace.json resolution path
Address review feedback on the community marketplace.json install path:
- Wrap PluginResolver.resolve() in try/catch so a malformed plugin entry
falls through to the legacy install path with a warn instead of
crashing cloneModule.
- Stop mutating the resolver's return object; shallow-clone before
stamping community provenance so install state cannot leak back into
resolver-owned objects.
- Warn when _selectPluginForModule lands on the single-plugin fallback
with a name that doesn't match the registry code or module_definition
hint, so a misconfigured marketplace.json can't silently install the
wrong plugin.
- Add CommunityModuleManager.resolveFromCache() and call it from
OfficialModules.install() when the in-process plugin cache is empty,
so callers that reach install() without pre-cloning still get the
marketplace-aware path. Reuses an existing channel resolution when
present, otherwise synthesizes a stable-channel stub from the registry
entry plus the cached repo's HEAD.
- Align installFromResolution()'s returned versionInfo.version with
manifestEntry.version precedence (communityVersion || cloneRef || ...)
so downstream summaries match what was written to the manifest.
Tests: lint, format:check, lint:md, test:install (290), test:channels
(83), test:refs (7) all green.
* fix(installer): resolve url-source custom modules from custom-modules cache
resolveInstalledModuleYaml previously only searched ~/.bmad/cache/external-modules/,
so modules installed via --custom-source <git-url> (cached at
~/.bmad/cache/custom-modules/<host>/<owner>/<repo>/) could not be located on
re-install runs. This caused warnings during npx bmad-method install:
[warn] collectAgentsFromModuleYaml: could not locate module.yaml for '<name>'
[warn] writeCentralConfig: could not locate module.yaml for '<name>'
Adds a fallback that walks the custom-modules cache via _findCacheRepoRoots
(identifying repo roots by .bmad-source.json or .claude-plugin/, not
marketplace.json, so direct-mode modules are also covered), reuses the same
searchRoot candidate-path logic, and matches by the discovered yaml's code
or name field.
Works without needing _resolutionCache to be populated, which fixes the
re-install scenario where no --custom-source flag is passed.
Closes#2312
* fix(installer): enumerate all module.yamls when walking custom-modules cache
A url-source custom-modules repo can host multiple plugins in discovery
mode (e.g. skills/module-a/module.yaml and skills/module-b/module.yaml).
The previous walk used searchRoot which returned only the first match,
so asking for module-b would surface module-a's yaml, fail the code/name
check, and skip the repo entirely — never inspecting module-b.
Splits the candidate-path traversal into searchRootAll (returns every
module.yaml in priority order) and a thin searchRoot wrapper for the
existing single-module fallbacks. The custom-modules walk now iterates
every yaml per repo and matches each against code or name.
- resolveInstalledModuleYaml: fall back to CustomModuleManager._resolutionCache for local
custom-source modules (external cache path doesn't exist for these); refactor candidate-path
search into shared searchRoot() helper; add *-setup/assets/module.yaml BMB standard path
- manifest-generator: use module code field (not display name) as TOML section key [modules.X]
Co-authored-by: cidemaxio <cidemaxio@users.noreply.github.com>
* fix(installer): mirror launch channel as default for external modules
When the user runs `npx bmad-method@next install`, the installer itself
runs from a prerelease, but the interactive channel gate previously hardcoded
"(all stable)" — defaulting tea/community modules to stable while bmad-method
itself was on next. The bleeding-edge launch did not flow through.
Detect the installer's own version via semver.prerelease() and default the
gate (and per-module picker) to match — "all next" for prerelease launches,
"all stable" for stable. Users keep full control: hit "n" to customize per
module, or pass explicit --channel / --pin / --next flags to override.
* fix(installer): seed channelOptions before module picker, not gate
CodeRabbit caught a label/install mismatch in the previous approach: the
module picker resolves version labels via decideChannelForModule, which runs
before _interactiveChannelGate. With channelOptions.global still null at
picker time, labels rendered from stable tags — then the gate flipped global
to 'next' and externals installed from main HEAD. Net effect on @next launches:
"tea (v1.6.0)" in the picker, but install pulled HEAD.
Move the launch detection up into promptInstall, immediately after
parseChannelOptions. Seeding channelOptions.global = 'next' before the picker
makes labels resolve from main HEAD (matching the install) and lets the
existing gate's haveFlagIntent check skip cleanly — the @next user already
declared their intent by typing it. Per-module customization remains available
via --pin / --next / --channel flags, same as for any pre-set global.
* fix(publish): advance @next dist-tag after stable release
When a stable release publishes via workflow_dispatch, @latest can leapfrog
the existing @next prerelease (e.g. latest=6.5.0 while next=6.4.1-next.0),
turning `npx bmad-method@next install` into a silent downgrade until the
next qualifying push to main republishes a fresh -next.0.
- publish.yaml: after stable publish, repoint @next at the just-published
stable version. The existing derive-prerelease step picks max(latest, next)
as its base, so subsequent push-driven prereleases bump from there.
- bmad-cli.js: checkForUpdate was querying the @beta dist-tag (which this
package does not use). Replace string-matching with semver.prerelease()
and query @next for prerelease users.
* fix(publish): harden next-tag advance step and broaden path filter
- continue-on-error on the dist-tag advance: failure leaves @next stale
until the next push-driven prerelease, which is recoverable; failing the
job after a successful publish + git tag + GH release is not.
- Status echo so release-log triage can confirm the advance ran.
- Add removals.txt to the push-trigger path filter. Installer-affecting
changes outside src/** (like the post-6.5.0 removals.txt fix) should
still trigger a fresh -next.0 publish.
Adds 32 entries to removals.txt covering the module-prefixed wrapper
skill names used pre-v6.2.0 (bmad-bmm-* and bmad-agent-bmm-*). Users
upgrading from v6.0.x / v6.1.x had these installed in their IDE skill
directories, but the v6.2.0 architecture switch dropped the module
prefix and the cleanup never knew the old names. Stale wrappers stayed
behind alongside the new self-contained skills, causing duplicates and
broken-file errors when invoked (referenced files no longer exist).
The removals.txt entries get added to the cleanup removalSet on every
install/update, so the next install run for an upgrading user removes
the stale wrappers automatically.
* refactor(installer): replace legacy_targets auto-cleanup with upgrade warnings
Removes the legacy_targets YAML field and its install-time auto-migration
of pre-v6.1.0 directories (.claude/commands, .opencode/agents, etc.). On
install, surface a warning instead: read manifest version and scan 24
known legacy paths, then print rm -rf commands the user can run themselves.
Also deletes orphan tools/platform-codes.yaml (never loaded by any code)
and fixes a stale URL in the cs translation.
* feat(installer): consolidate to .agents/skills and add global_target_dir for all platforms
Updates platform-codes.yaml against verified primary docs for all 24 supported
platforms. 14 platforms (auggie, codex, crush, cursor, gemini, github-copilot,
kilo, kimi-code, opencode, pi, roo, rovo-dev, windsurf) move their project
target_dir to the cross-tool .agents/skills/ standard. Junie moves from the
broken .agents/skills/ to its own .junie/skills/ per JetBrains docs.
Adds global_target_dir to every platform: 11 share ~/.agents/skills/, Crush
uses XDG ~/.config/agents/skills/, Codex global stays ~/.codex/skills/, the
rest are tool-specific. Ona and Trae omit global (no documented home path).
Note: installer logic does not yet dedupe writes for platforms sharing a
target_dir — users installing multiple .agents/skills/ tools together will
overwrite the same files (harmless on install, but uninstalling one clears
the dir for the others). Coordination logic is the next step.
* feat(installer): add 18 new platforms, dedup shared target_dir, ownership-aware cleanup
Adds 18 platforms from the verified Vercel list (adal, amp, bob, command-code,
cortex, droid, firebender, goose, kode, mistral-vibe, mux, neovate, openclaw,
openhands, pochi, replit, warp, zencoder). Marks codex and github-copilot as
preferred alongside claude-code and cursor.
Coordination for platforms sharing a target_dir:
- IdeManager.setupBatch dedups skill writes when multiple selected platforms
point at the same target_dir (e.g. .agents/skills/). The first platform
writes, peers skip the redundant wipe-and-rewrite. Result reports the same
count and target dir for every member so the install summary is consistent.
- IdeManager.cleanupByList accepts remainingIdes; when removing one platform
from a shared dir while another co-installed platform still owns it, the
target_dir wipe is skipped. Platform-specific hooks (copilot markers, kilo
modes, rovodev prompts) still run.
- _setupIdes uses setupBatch; _removeDeselectedIdes passes remainingIdes so
partial reconfigure preserves shared skills.
Skill ownership now uses skill-manifest.csv canonicalIds, not the bmad- prefix.
This unblocks custom modules that ship skills with non-bmad names (e.g.
fred-cool-skill). Affected sites:
- _config-driven.detect: reads canonicalIds from the project's bmadDir
- _config-driven.findAncestorConflict: reads canonicalIds from the ancestor's
own bmadDir, falling back to the prefix only when no manifest exists
- legacy-warnings.findStaleLegacyDirs: same canonicalId-based detection
Migration warnings: LEGACY_SKILL_PATHS adds 12 skill dirs that moved to the
.agents/skills/ standard (cursor, gemini, github-copilot, kimi, opencode, pi,
roo, rovodev, windsurf, plus their globals). Users with stale skills in those
locations get a one-line warning with the rm command per dir.
New shared helper tools/installer/ide/shared/installed-skills.js exposes
getInstalledCanonicalIds(bmadDir) and isBmadOwnedEntry(entry, canonicalIds).
Tests: 9 new assertions across two suites covering dedup, partial uninstall
preservation, and custom-module skill detection. All 286 tests pass.
* fix(installer): setupBatch must not claim a shared target_dir on failure
If the first platform's setup throws or returns success: false, the dedup map
previously still recorded the claim with skillCount: 0, causing every peer
sharing the target_dir to skip its install — leaving the dir empty/broken
behind a cascade of misleading "shares with X" rows.
Now the claim is only recorded when the install succeeded and wrote skills.
On failure, the next peer becomes the new first writer and recovers.
Adds Suite 40b regression test that monkey-patches cursor.setup to throw
and verifies gemini still populates the shared dir.
* fix(installer): address PR #2313 review findings
Three issues raised by augmentcode and coderabbit bot reviewers:
1. _removeDeselectedIdes silently swallowed cleanup failures after the
refactor to cleanupByList. The old per-IDE try/catch logged a warning;
the new path discarded the result array. Now logs a warning per failed
ide so failures stay visible.
2. The legacy-dir cleanup hint printed `rm -rf "<path>"/bmad*` which both
matched bmad-os-* utility skills the user should keep AND missed the
custom-module skills (e.g. fred-cool-skill) that the new canonical-id
detection now finds. Findings now carry the exact entry names from the
scan, and the warning prints one precise rm line per entry.
3. warnPreNativeSkillsLegacy did unguarded fs reads at install start. A
permission/IO error would have aborted the whole install. Wrapped the
call site in try/catch so legacy-scan failures only emit a warning.
Adds kimi-code to both platform-codes.yaml files so Kimi Code CLI
is available as an install target via the config-driven installer.
Skills are installed to .kimi/skills/, which is the project-level
skills directory per the official Kimi Code CLI documentation.
Closes#1630
Co-authored-by: Brian <bmadcode@gmail.com>
When a story modifies existing files, create-story must read those
files before generating dev notes. Without this, dev agents improvise
design decisions without knowing the current state of the code, leading
to regressions caught only at review time.
Adds a step at the end of Step 3 (Architecture analysis) that reads
every file marked UPDATE in the architecture directory structure and
documents its current state, what the story changes, and what must
be preserved.
Fixes#2273
Co-authored-by: Brian Madison <bmadcode@gmail.com>
* feat: extend customize.toml to all 6 developer-execution workflows (#2303)
Add uniform customization support to dev-story, code-review,
sprint-planning, sprint-status, quick-dev, and checkpoint-preview,
matching the same 4 extension points (activation_steps_prepend,
activation_steps_append, persistent_facts, on_complete) already
available on 17 BMM workflows from PR #2287.
- Create customize.toml for each workflow
- Add 6-step activation block to SKILL.md (merge workflow.md
content in, delete workflow.md per PR #2287 pattern)
- Wire on_complete at terminal steps (inline <action> for XML
workflows, ## On Complete section for step-file workflows)
- Fix pre-existing step number reference in dev-story (Step 6 → 9)
* fix: correct goto step="6" → step="9" in dev-story
The XML goto at line 203 still pointed to step 6 ("Author
comprehensive tests") instead of step 9 ("Story completion and mark
for review"), which is the actual completion gate. This was the
same class of pre-existing bug fixed in the text (M-1) but
missed in the XML action.
---------
Co-authored-by: Brian <bmadcode@gmail.com>
* feat(installer): channel-based version resolution for external modules
Adds stable/next/pinned channel resolution so external/community modules
install at released git tags by default instead of tracking main HEAD.
Manifest now records channel, resolved version, and SHA per module for
reproducible installs.
CLI flags: --channel, --all-stable, --all-next, --next=CODE (repeatable),
--pin CODE=TAG (repeatable). Precedence: pin > next > channel > registry
default > stable. --yes accepts patch/minor upgrades but refuses majors.
Interactive "Ready to install (all stable)?" gate with a per-module
picker (stable/next/pin) when declined. Re-install prompts classify tag
diffs as patch/minor/major with semver-class-dependent defaults.
Legacy version:null manifests get a one-time migration prompt.
Custom modules gain an optional @<ref> URL suffix for pinning (https,
ssh, /tree/<ref>/subdir forms supported; local paths rejected).
Community modules honor --next/--pin overrides with a curator-bypass
warning; default path still enforces the approved SHA.
Quick-update now reads the manifest's recorded channel per module so
pinned installs don't silently roll forward.
* feat(installer): interactive channel switch, upgrade refusal, unified docs
Builds on the channel-resolution foundation. The installer now lets users
flip a module between stable, next, and pinned after install — either
interactively via a "Review channel assignments?" gate, or by flag. Quick
and modify re-installs classify stable upgrades; under non-interactive
flows, patches and minors apply automatically but majors are refused with
a pointer to --pin.
Fallback behavior for GitHub rate-limit / network failures is now cache-
aware: re-installs reuse the recorded ref silently; fresh installs abort
with actionable guidance (set GITHUB_TOKEN or use --next/--pin). Bundled
modules (core, bmm) warn when targeted by --pin or --next so users aren't
left wondering why the flag had no effect.
Install summary labels no longer mangle "main" into "vmain"; next-channel
entries render as "main @ <short-sha>" instead. Bundled modules are now
correctly skipped from all channel prompts and tag-API lookups.
Docs consolidated into a single how-to. install-bmad.md now covers the
interactive flow, the channel model (stable/next/pinned plus the npm
dist-tag axis for core/bmm), the re-install upgrade prompts, the full
flag reference, copy-paste recipes, and troubleshooting. The old
non-interactive-installation.md is reduced to a redirect stub.
* fix(installer): review fixes + unit tests for channel resolution
- ui.js: import parseGitHubRepo; fixes ReferenceError in the
interactive channel picker's stable-tag pre-resolve path.
- community-manager: pinned modules now fetch+checkout the pin tag
on cache refresh instead of resetting to origin/HEAD (was silently
drifting to main on re-install).
- channel-plan: parseChannelOptions returns acceptBypass so --yes
auto-confirms the curator-bypass prompt; headless --next/--pin
installs of community modules no longer hang.
- community-manager: simplify recordedVersion (dead ternary branch).
- custom-module-manager: drop "or sha" from the @<ref> comment
(git clone --branch rejects raw SHAs); update-path fetches
origin <ref> so /tree/<branch>/ URLs work too.
- install-bmad.md: rename "Headless / CI installs" to "Headless CI
installs" so the stub's #headless-ci-installs anchor resolves.
- test/test-installer-channels.js: 83 unit tests for channel-plan
and channel-resolver pure modules; wired into npm test as
test:channels.
* fix(installer): address CodeRabbit review findings
- ui.js: skip stable-channel upgrade classification when the user has
already declared intent via --pin/--next=/--channel or the review
gate. Prevents the decline / major-refused / fetch-error branches
from silently overwriting an explicit pin with prev.version.
- external-manager.js: short-circuit cloneExternalModule when the
requested plan matches an existing in-process resolution and the
cache is valid. Avoids redundant resolveChannel() + git fetch on
every same-plan lookup in a single install.
- installer.js: fall back to CommunityModuleManager.getResolution()
when no external resolution exists, so community module result
rows carry newChannel/newSha instead of null under --next/--pin.
- installer.js: don't label a module as "no change" when its version
string is 'main'/'HEAD' — the SHA may have moved and preVersions
doesn't track the prior SHA. Show "(refreshed)" instead.
- official-modules.js: match versionInfo.version to the manifest's
cloneRef || (hasGitClone ? 'main' : version) expression so summary
lines report the cloned ref for git-backed custom installs.
- install-bmad.md: clarify that sha is only written for git-backed
modules and that rerunning the same --modules on another machine
does not reproduce stable-channel installs — convert recorded tags
into explicit --pin flags for cross-machine reproducibility.
External official modules (bmb, cis, gds, tea, wds) are cloned to
~/.bmad/cache/external-modules/<name>/ and never copied into src/modules/,
so collectAgentsFromModuleYaml silently skipped them and their agents
never reached config.toml. Swap the hardcoded src/modules lookup for a
resolveInstalledModuleYaml() helper that also searches the external cache
(handling src/, skills/, nested, and root layouts) and warns instead of
silently skipping when a module.yaml can't be found.
All 16 bare workflow customize.toml files now have the same thorough comment
block as the well-documented ones, with skill-specific on_complete descriptions
that name the exact terminal step and exit condition.
on_complete is now executed at the true end of each workflow's terminal step
rather than lazily referenced in SKILL.md:
- Linear workflows: ## On Complete block appended to the final step file
(create-prd step-12, create-ux-design step-14, create-architecture step-08,
generate-project-context step-03, check-implementation-readiness step-06,
epics-and-stories step-04, all three research step-06 files, prfaq verdict,
document-project both sub-workflow instruction files)
- Multi-path workflows: on_complete inline on each true exit path only
(edit-prd fires on [S] Summary and [X] Exit, not on [V] Validate or [E] Edit;
validate-prd fires on [X] Exit only, not on [R], [E], or [F])
- Inline XML workflows: <action> tag at the close of the final step
(correct-course step-6, create-story step-6, retrospective step-12,
qa-generate-e2e-tests appended to SKILL.md)
* feat(core-skills): add bmad-customize for authoring _bmad/custom overrides
A conversational guide skill that helps users author or update TOML overrides
in _bmad/custom/ for customizable BMad agents and workflows. Covers per-skill
agent and workflow surfaces; central config is out of scope for v1.
- SKILL.md: six-step flow (intent, discover, route, compose, team-vs-user,
show-confirm-write-verify) with baked-in agent-vs-workflow routing heuristic
and a template-swap subroutine
- scripts/list_customizable_skills.py: stdlib-only scanner that enumerates
customizable skills across standard IDE install paths, reports surface type
and override status, PEP 723, 10 unit tests
- Reuses _bmad/scripts/resolve_customization.py for post-write verification
- Registered in core-skills/module-help.csv with menu code BC
* refactor(bmad-customize): apply QA pass (top 3 recommendations)
Applies the three highest-payoff themes from the quality analysis:
- Labeling + completion contracts: rename ## Purpose to ## Overview,
add domain framing (what customization means in BMad, typical user
arrival shapes), add an explicit Completion block with testable
conditions for "skill run is done"
- Hostile-environment robustness: add On-Activation preflight that
classifies no-BMad / BMad-without-resolver / full-install states,
instruct Step 2 to surface scanner errors[] and scanned_roots on
empty results, add resolver-missing fallback to Step 6.4, add a
re-enter-Step-4 recovery loop when verify shows the override didn't
take effect
- Returning-user and iteration experience: add "Audit / iterate"
intent class in Step 1, lead discovery with already-overridden
skills for that intent, read existing overrides in Step 3 before
composing, frame Step 4 as additive-on-top rather than fresh
authoring, give Cross-cutting intent an explicit Step 3 branch
that walks agent-vs-workflow with the user
Resolves 12 of 18 observations from the quality report. Lint clean
(scan-path-standards and scan-scripts both 0 findings). Unit tests
still 10/10.
* refactor(bmad-customize): derive skills root from install location
Previously the scanner hardcoded a list of IDE skill directories
(.claude/skills, .cursor/skills, .cline/skills, .continue/skills) and
scanned them relative to the project root. That was wrong: skills can
be installed either project-local or user-global, the IDE determines
the convention, and the set of valid locations is open-ended.
The scanner now derives its primary skills root from __file__ — the
running skill's own install directory is the authoritative location
for finding siblings. --skills-root overrides the default; --extra-root
(repeatable) adds additional locations for the rare mixed-install case.
Changes:
- list_customizable_skills.py: remove SKILL_ROOTS constant, add
default_skills_root() derived from __file__, rename scan_project
to scan_skills(skills_roots, project_root), add --skills-root and
--extra-root flags, de-dupe skills when the same name appears in
multiple roots (first wins)
- SKILL.md: update Step 2 to describe the scanner's derive-from-install
behavior and when to use --extra-root; drop the hardcoded IDE path
list from Notes
- tests: refactor setUp to place skills under a generic skills root
(not .claude/skills), add 3 new tests for multiple-roots merge,
duplicate-name precedence, and missing-root error reporting
* docs(customization): point users at bmad-customize as the guided path
Surface the new bmad-customize skill across the three customization
docs so users know they don't need to hand-author TOML to benefit
from the surface:
- customize-bmad.md: prominent tip at the top introducing the skill
as the guided authoring helper; updated the "Need to see what's
customizable?" troubleshooting tip to recommend the skill first
- expand-bmad-for-your-org.md: tip under prereqs noting every recipe
can be applied via the skill, with the recipes remaining the
reference for what to override
- named-agents.md: short paragraph in the customization section and a
link entry under the references list
Hand-authoring still works the same way; the skill is additive.
Central-config overrides are flagged as the current exception.
* docs(bmad-customize): steer users at bmad-builder instead of 'forking'
* fix(bmad-customize): reword description to pass file-ref validator
* refactor(bmad-customize): tighten description and expand module-help entry
- SKILL.md description: drop the catch-all 'or asks how to change the
behavior of a specific BMad skill' trigger clause that would fire in
casual discussion; keep the four explicit phrase triggers.
- module-help.csv: rewrite the description so bmad-help has real
routing material — names the concrete capabilities (persistent
facts, template swaps, activation hooks, menus), the scope routing,
and the value prop (no TOML hand-authoring). Matches the 'Use
when...' pattern other Core entries use.
* fix(module-help): quote bmad-customize description field that contains commas
* fix(bmad-customize): address PR #2289 review findings
- SKILL.md preflight: load root config from _bmad/config.toml and
config.user.toml (not .yaml) — the installer emits TOML; the YAML
references would have made the skill silently miss real user config
- SKILL.md resolver fallback (Step 6.4): read all three merge layers
when present (base / team / user) and describe the merge in
base → team → user order; the prior wording could describe the wrong
effective merge when the user wrote .user.toml on top of an existing
team .toml
- SKILL.md: replace bare 'docs/how-to/customize-bmad.md' references
(3 locations) with the public docs URL so users installing the skill
aren't pointed at a path they don't have locally
- list_customizable_skills.py: catch UnicodeDecodeError in
read_frontmatter_description so a non-UTF-8 SKILL.md can't abort
the whole scan
- list_customizable_skills.py: clarify exit-code contract in the
module docstring — errors[] is non-fatal by design, exit 2 is
reserved for invocation errors
- customize-bmad.md: tighten the tip to scope bmad-customize to the
per-skill surface; central-config is out of scope v1
- expand-bmad-for-your-org.md: same scoping — Recipes 1-4 can be
applied by the skill; Recipe 5 (central config) stays hand-authored
* fix(bmad-customize): markdownlint MD034 and validate-file-refs
- Wrap the three docs.bmad-method.org references as
[text](url) markdown links instead of bare URLs (MD034)
- Drop the {project-root}/ prefix on line 41's config.toml
references. validate-file-refs strips the template prefix and
tries to resolve 'config.toml' as 'src/config.toml'; sibling
skills (party-mode, retrospective, advanced-elicitation) all
reference '_bmad/config.toml' bare and pass CI — match that
pattern. The '(root level under {project-root}, installer-owned)'
parenthetical preserves the disambiguation.
* refactor(bmad-customize): cut token-wasting prose from SKILL.md
Down from 175 lines to 110. Removed:
- 'What customization means in BMad' architecture backgrounder — the
LLM reads the live customize.toml in Step 3; doesn't need the lore
- 'Desired Outcomes' section — retrospective narration of what the
6 steps already instruct
- 'Role' section — fluff; the flow itself defines the role
- 'Notes' section — sparse-override rule already in Step 4, IDE-path
note is commentary, docs link duplicates the out-of-scope section
- 'The scanner derives its skills directory from...' and 'returns JSON
with...' — commentary the LLM doesn't need; it runs the script and
sees the output
- 'that file IS the schema' and similar editorial asides throughout
- Explanatory clauses like 'silently drifts on every release' and
'trust the user's domain knowledge'
Kept everything that's load-bearing: preflight conditionals, intent
classification, routing heuristic, merge semantics, template-swap
subroutine, team-vs-user defaults, verify fallback and recovery loop,
completion conditions, out-of-scope list.
* feat(skills): add TOML workflow customization to 17 bmm-skills
Flattens each skill's workflow.md into SKILL.md and adds a customize.toml
surface with a 6-step activation block (resolve_customization, prepend,
persistent_facts, config, greet, append). Core-skills and developer
execution skills (dev-story, code-review, sprint-planning, sprint-status,
quick-dev, checkpoint-preview) are intentionally excluded.
Customized: document-project, prfaq, domain/market/technical-research,
create-prd, create-ux-design, edit-prd, validate-prd,
check-implementation-readiness, create-architecture,
create-epics-and-stories, generate-project-context, correct-course,
create-story, qa-generate-e2e-tests, retrospective.
* fix(skills): address PR review findings on workflow customization
- bmad-create-story: drop stale {project_context} variable reference
from step 2 note; content is already loaded via persistent_facts
- research skills (market/domain/technical): derive research_topic_slug
before writing output filename to prevent path injection and invalid
filesystem characters
- bmad-correct-course: reconcile step 1 verify list and HALT with the
"Missing documents" rule — Architecture and UI/UX are optional, HALT
only fires when PRD or Epics are missing
- fix grammar in Micro-file Design bullets across 5 migrated skills
(self contained → self-contained, adhere too 1 file → adhere to one
file at a time)
- docs/how-to/customize-bmad.md: document workflow activation order
and frame the baseline fields as a stable initial pass with targeted
per-workflow customization points coming later
* feat(agents): set team to software-development on BMM agents
All six BMM agents (analyst, tech-writer, PM, UX designer, architect,
dev) now explicitly declare `team: software-development` in the
module.yaml roster instead of falling back to the module-code default
of `bmm`.
This matches the BMad-wide team convention where agents across modules
that collaborate on software delivery share one named team. Tea's Murat
joins the same team via a parallel PR in bmad-method-test-architecture-
enterprise so party-mode, help catalog, and retrospective skills can
route the full software-delivery roster as a single unit.
* test: update team assertions for explicit software-development
* refactor: remove bmad-skill-manifest yaml; introduce four-layer central config.toml
- Agent essence moves from per-skill bmad-skill-manifest.yaml files
into each module.yaml's `agents:` block (code, name, title, icon,
description). Per-agent customize.toml remains the deep-behavior
source of truth.
- Installer emits four TOML files:
_bmad/config.toml team install answers + agent roster
_bmad/config.user.toml user install answers
_bmad/custom/config.toml team overrides stub
_bmad/custom/config.user.toml personal overrides stub
Prompts declare scope: user to route answers to config.user.toml.
- resolve_config.py merges four layers: base-team -> base-user ->
custom-team -> custom-user.
- Three consumer skills (party-mode, advanced-elicitation,
retrospective) switched from agent-manifest.csv to the resolver.
- installer.js mergeModuleHelpCatalogs now takes the in-memory
agent list from ManifestGenerator -- no CSV roundtrip.
- Deleted: 6 bmad-skill-manifest.yaml files, agent-manifest.csv
emission, collectAgents/getAgentsFromDirRecursive,
paths.agentManifest().
* fix(installer): strip core-key pollution from [modules.*]; soften config headers
- writeCentralConfig now always strips core-module keys from every
[modules.<code>] bucket, even when the module's schema is not
available in src/ (external / marketplace modules like cis, bmb).
Core values belong in [core] only; workflows read them directly.
- When the module's own schema IS available (built-in modules),
also drop any key it does not declare as a prompt — same
spread-pollution filter as before, now layered on top.
- Section-aware headers on both _bmad/config.toml and
_bmad/config.user.toml: [core] / [modules.*] values are
editable (installer reads them as defaults on next install);
[agents.*] is regenerated from module.yaml and will be wiped —
overrides for agents go in _bmad/custom/config*.toml instead.
* docs: cover central config.toml + Diataxis prose pass across three files
Document the new four-file central configuration surface (_bmad/config.toml,
config.user.toml, and custom/ overrides) alongside the existing per-skill
customize.toml. Make editing rules, scope partitioning, and when-to-use-which
guidance explicit.
- customize-bmad.md: new "Central Configuration" section with editing rules,
three worked examples (rebrand, fictional agent, module settings override),
and a "when to use which surface" table. Converted five h4 headers to
bold paragraph intros per style guide.
- expand-bmad-for-your-org.md: two-layer mental model extended to three;
new Recipe 5 with three variants (rebrand, custom crew, pinned team
settings); reinforcement table extended.
- named-agents.md: noted the dual customization surface — per-skill shapes
behavior, central config shapes roster identity.
Diataxis prose pass applied across all three files: banned vocabulary
check, em-dash cap, hypophora / metanoia / amplificatio / stakes-inflation
cleanup, rhythm and burstiness fixes. Structural conformance verified;
markdownlint and prettier clean.
* test+docs: add central config unit tests; fix stale recipe count
- test: two new suites (35 + 36) covering writeCentralConfig and
ensureCustomConfigStubs. Verifies scope partitioning (user_name
lands only in config.user.toml), core-key pollution stripping
from [modules.*], unknown-schema fallthrough (external modules
survive without schema), agent roster baked into config.toml
[agents.*] only, stub-preservation on re-install. 44 new
assertions.
- docs: fixed four stale "four recipes" references to say "five"
after Recipe 5 (Customize the Agent Roster) was added. Touches
frontmatter, opening paragraph, Combining Recipes paragraph,
and the named-agents cross-link blurb.
* fix: address PR review feedback on central config
- resolve_config.py argparse: three-layer → four-layer description
- SKILL/workflow/explanation docs: document all four layers including
_bmad/config.user.toml (was missing from merge-stack descriptions)
- customize-bmad.md + installer headers: drop the false "direct edits to
config.toml persist" claim; installer reads from per-module config.yaml,
not central TOML, so direct edits get clobbered. Route users to
_bmad/custom/config.toml for durable overrides
- writeCentralConfig: warn loudly when a module.yaml can't be parsed
(previously silent — user-scoped keys could mis-file into team config)
- writeCentralConfig: preserve [agents.*] blocks for modules that didn't
contribute fresh agents this run (e.g. quickUpdate skipping modules
whose source is unavailable) so the roster doesn't silently shrink
- add extractAgentBlocks helper + Test Suite 37 covering preservation
Addresses comments from augmentcode and coderabbitai on PR #2285.
* feat(skills): TOML-based agent customization with stdlib Python resolver
Re-applies PR #2282's three-layer customization model (skill defaults →
team → user) but swaps YAML for TOML and uv for stdlib tomllib. Users
no longer need uv, pip, or a virtualenv — plain python3 (3.11+) is
sufficient, since tomllib shipped in the standard library.
## Schema changes vs PR #2282
- Flat agent schema: fields live directly under [agent], no nested
metadata/persona sub-tables. Easier to author, less indentation.
- Non-configurable identity: name and title are declared in
customize.toml as source-of-truth metadata (for future skill-manifest
generation) but SKILL.md ignores overrides there — identity is
hardcoded to preserve brand recognition.
- role redefined: now describes what the skill does for the user
within its module phase, not a restatement of the title.
- persistent_facts replaces the activation-time file-context load AND
the old memories concept. Entries can be literal sentences or
file: prefixed paths/globs; avoids collision with the upcoming
runtime memory sidecar.
- activation_steps_prepend / activation_steps_append harmonized across
agents and workflows (replaces agent-specific critical_actions).
- [workflow] namespace mirrors [agent] for workflow customization.
Same four structural rules, same field vocabulary.
## Resolver (src/scripts/resolve_customization.py)
Four purely structural merge rules, zero field-name hardcoding:
- Scalars: override wins
- Tables: deep merge
- Arrays of tables where every item has `code` or `id`: merge by
that key (matching keys replace, new keys append)
- Any other array: append
No removal mechanism — overrides cannot delete base items. Fork the
skill or override by code with a no-op value to suppress defaults.
## Agents ported (6)
All six BMad agents now ship customize.toml + rewritten SKILL.md:
analyst (Mary), tech-writer (Paige), pm (John), ux-designer (Sally),
architect (Winston), dev (Amelia). Each uses the same 8-step
activation template: resolve → execute prepend → adopt persona →
load persistent facts → load config → greet (with {agent.icon}) →
execute append → dispatch or present menu.
Step 8 supports fast-path invocation: "hey Mary, let's brainstorm"
dispatches the matching menu item directly after greeting, skipping
the menu render when intent is clear. Chat, clarifying questions,
and bmad-help remain available when nothing on the menu fits.
## Installer + tooling
- _bmad/scripts/ provisioned on install (copies src/scripts/)
- _bmad/custom/ seeded with .gitignore for *.user.toml on fresh install
- Non-module-dir filter extended to skip _memory, memory, docs,
scripts, and custom when scanning for modules
- Dead _config/agents/ directory no longer created
- metadata.capabilities removed from agent-manifest.csv and schema
- eslint config extended to cover src/scripts/**
- validate-file-refs.js knows about custom/ as install-only
## Deferred for follow-up
- bmad-product-brief workflow port (the pilot that demonstrates
[workflow] + on_complete)
- Translated docs (cs/fr/vi-vn/zh-cn) — regenerate from English
* feat(skills): port bmad-product-brief to TOML workflow customization
Completes the customization surface rollout by giving the product-brief
workflow the same override model as the six BMad agents, under the
[workflow] namespace instead of [agent].
## customize.toml
Mirrors the agent shape under [workflow] with:
- activation_steps_prepend / activation_steps_append (harmonized across
agents and workflows — same field names, same append semantics)
- persistent_facts with the file: convention, seeded with
file:{project-root}/**/project-context.md
- on_complete scalar (renamed from PR #2282's skill_end for clarity —
reads cleaner as "what runs when the workflow completes")
## SKILL.md
7-step workflow activation:
1. Resolve workflow block
2. Execute prepend steps
3. Load persistent facts (file: or literal)
4. Load config
5. Greet if not already
6. Execute append steps
7. Stage 1 — Understand Intent
python3 + stdlib tomllib invocation; no uv required.
## Prompt file changes
- Path normalization: ../agents/ → agents/, ../resources/ → resources/,
bare foo.md → prompts/foo.md. All references now resolve from the
skill root (matches the convention documented in SKILL.md).
- Paths: meta-line added to each of the 4 prompt files that reference
other files, reinforcing "bare paths resolve from skill root" so the
LLM doesn't lose the convention when operating two hops into a
prompt chain.
- finalize.md terminal stage now calls the resolver for
workflow.on_complete — non-empty values run as the final step.
## Validation
- Resolver output verified: 4 workflow fields returned cleanly.
- validate-file-refs.js: 254 files scanned, 139 refs checked, 0 broken.
- test:refs: passing.
* docs(skills): enterprise customization recipes + workflow template variable
Three independent improvements bundled because they share the same
surface (workflow/agent customization) and landed from the same design
discussion:
## Fallback sentence disambiguated (7 SKILL.md files)
The "if the script fails" fallback used to say `{project-root}/_bmad/
custom/{skill-name}.toml` for the team override and then just `{skill-
name}.user.toml` for the user override, leaving the user file's
location implicit. LLMs could reasonably guess skill root or project
root instead. Replaced with an unambiguous numbered list that spells
out the full path for every file in the merge chain.
## Product-brief: stage promotion + brief_template variable
- Promoted `## Stage 1: Understand Intent` from a nested step inside
"On Activation" to a top-level section. The previous "Step 7: Stage
1 — Understand Intent → Proceed to Stage 1 below" was mechanical
numbering pretending to be a step. Activation now ends cleanly at
Step 6; Stage 1 is a peer section.
- Added `brief_template` as a workflow-level scalar customization
defaulting to `resources/brief-template.md`. Stage 4 reads
`{workflow.brief_template}` instead of the hardcoded path, so orgs
can point at their own template under `{project-root}/...` without
forking the skill.
## New doc: docs/how-to/extend-bmad-for-your-org.md
Four worked recipes that together cover most enterprise scenarios:
1. Shape an agent across every workflow it dispatches (dev agent +
Context7 MCP + Linear search — the highest-leverage pattern)
2. Enforce org conventions inside a specific workflow (product-brief
+ compliance-field persistent_facts)
3. Publish completed outputs to external systems (product-brief +
Confluence + Jira via MCP, gated on user confirmation for Jira)
4. Swap in your own output template (product-brief + brief_template
variable swap)
Opens with the two-layer mental model (agent spans workflows,
workflow is local) so readers pick the right granularity before
reading any recipe. Closes with a "Combining Recipes" section
showing all four composed. Cross-linked from customize-bmad.md.
## Validation
- Resolver: workflow.brief_template returns the default cleanly.
- validate-file-refs.js: 254 files scanned, 146 refs checked
(+7 from this commit), 0 broken.
* docs(skills): encourage CLAUDE.md/AGENTS.md reinforcement of critical rules
Added a "Reinforce Global Rules in Your IDE's Session File" section to
extend-bmad-for-your-org.md. BMad customizations only load when a
skill activates, but IDE session files (CLAUDE.md, AGENTS.md, cursor
rules, copilot-instructions) load every turn — worth restating the
most critical rules there too so they survive ad-hoc chat outside a
BMad skill.
Includes a one-line example reinforcing the Recipe 1 Context7 rule,
plus a scope table that clarifies what each layer is for:
- IDE session file: universal, every session, keep succinct
- Agent customization: persona-specific, every dispatched workflow
- Workflow customization: one workflow run
Emphasizes brevity — noise in the session file crowds out signal.
* docs(skills): add Named Agents explanation doc
New docs/explanation/named-agents.md walking through the three-legged
stool (skills + named agents + customization) with the "Hey Mary,
let's brainstorm" activation flow as the narrative thread.
Covers:
- Why named agents vs menu-driven or prompt-driven alternatives
- The 8-step activation flow and what each step contributes
- How customization scales the model beyond a single developer
- Cross-links to the how-to docs for implementation details
Sits alongside brainstorming.md, quick-dev.md, party-mode.md in the
explanation folder — feature narratives for users who want to
understand why BMad is designed the way it is, not just how to use it.
* docs(skills): clarify that keyed-merge requires a single identifier key per array
Review feedback (PR #2284) flagged that the merge-rules wording was
ambiguous: "every item has a `code` or `id` field" could reasonably
be read as "each item individually has at least one of the two",
allowing arrays to mix `code` and `id` across items.
The resolver has always required all items share the *same* identifier
key (all `code`, or all `id`). Mixed arrays fall through to append —
intentional, because mixing identifier keys within one array is a
schema smell and any guess about which key should merge creates a
worse trap than the append-fallback.
Clarified in three places:
- Merge-rules table in customize-bmad.md: "every item shares the
**same** identifier field"
- `code`/`id` convention paragraph: "pick **one** convention ... and
stick with it across the whole array"
- Resolver docstring and `_detect_keyed_merge_field` docstring:
explicit note that mixed arrays fall through with rationale
No behavior change.
* docs(skills): address CodeRabbit review — fallback rules, OS claim, headless greeting
Three fixes from PR #2284 review feedback:
## 1. Fallback merge wording (7 SKILL.md files)
Every SKILL.md told the LLM to merge the three customization files
"in priority order (later wins)" when the resolver fails. That reads
as shallow last-write-wins — but the resolver does structural merge
(scalars override, tables deep-merge, code/id-keyed arrays merge by
key, other arrays append). Following the old wording manually would
have silently stripped base `principles`, `persistent_facts`, and
`menu` items whenever a team override was present.
Expanded the fallback sentence to restate the four structural rules
explicitly, matching the resolver's behavior.
Applied to all 6 agents + bmad-product-brief workflow.
## 2. Python 3.11 / OS shipping claim (customize-bmad.md)
The docs claimed "macOS 13+, Ubuntu 22.04+, Debian 12+, Fedora 37+
all ship 3.11 or newer." Inaccurate — Ubuntu 22.04 defaults `python3`
to 3.10.6 (3.11 is a separate package), and macOS doesn't really
ship Python by default anymore.
Replaced with honest guidance: check `python3 --version` and note
that macOS without Homebrew and Ubuntu 22.04 default to 3.10 or
earlier.
## 3. Autonomous mode greeting gate (bmad-product-brief)
Product-brief's activation-mode detection documents autonomous mode
as "produce complete brief without interaction" — but Step 5 greeted
unconditionally, adding conversational output before the headless
artifact. Gated the greeting on `{mode}` != `autonomous`.
## Dismissed (replied on thread)
- `.gitignore` migration from *.user.yaml to *.user.toml: YAML
installer code was in reverted #2282, never released. No users
affected. Same rationale as Augment's earlier thread.
Validated: 254 files, 146 refs, 0 broken. test:refs 7/7,
test:install 242/242.
* docs: rename Extend to Expand throughout customization docs
Three-layer customization (skill defaults → team → user) for BMad agents
and any skill that opts in. Users edit `_bmad/custom/{skill-name}.yaml`
(team, committed) or `{skill-name}.user.yaml` (personal, gitignored);
customizations survive updates.
Resolver is a Python script using PEP 723 inline metadata, invoked via
`uv run` so deps auto-install into a cached isolated env on first call.
This aligns with Anthropic's Agent Skills spec and BMB conventions, and
keeps the dependency declared (scannable by pip-audit/Dependabot) rather
than vendored.
## Design choices
- **Agent identity is hardcoded** in SKILL.md (name, title, Overview prose)
so skills can be invoked reliably by role *or* default name. Brand
recognition is preserved; customization shapes behavior, not identity.
- **Luminary-anchored personas** (e.g. "Channels Martin Fowler's
pragmatism and Werner Vogels's cloud-scale realism") deliver ~55%
token savings per agent while preserving distinctive voice beats.
- **Universal per-field merge rules** with v6.1-compatible agent
semantics: metadata shallow-merge, persona replace, critical_actions
and memories append, menu merge-by-code, all else deep-merge.
- **Workflow customization** shares the same surface — `bmad-product-brief`
pilots `activation_steps_prepend`, `activation_steps_append`, and
`skill_end` hooks that any workflow-style skill can adopt.
## Infrastructure
- `_bmad/scripts/` houses shared Python scripts (resolver + future).
- `_bmad/custom/` is provisioned empty with a seeded `.gitignore` for
`*.user.yaml` on fresh installs.
- Installer filters ensure `scripts/`, `custom/`, and sidecar-generated
`memory/` directories are never treated as modules.
- Dead v6.1 code cleaned up: `_config/agents/` no longer created,
`metadata.capabilities` removed from schema and CSV manifest.
* feat(installer): use GitHub API as primary fetch with raw CDN fallback
Corporate proxies commonly block raw.githubusercontent.com while allowing
api.github.com. Add fetchGitHubFile() to RegistryClient that tries the
GitHub Contents API first, falling back to the raw CDN transparently.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(installer): cap redirect depth and preserve dual-fallback errors
Add maxRedirects parameter to fetch() and _fetchWithHeaders() to prevent
unbounded redirect recursion. Wrap CDN fallback in try/catch and throw
AggregateError with both API and CDN errors for better diagnostics.
Extract marketplace repo coordinates into named constants in
external-manager.
* chore(installer): drop unused fetchJson and fetchGitHubJson
Neither method has any callers. Also drop the corresponding test.
* refactor(test): fold registry tests into test-installation-components
No reason for RegistryClient tests to be a separate runner — the same
file already tests the registry consumers in Suite 33. Drop test:registry
from package.json scripts and quality gate.
* fix(installer): include URL, API message, and rate-limit info in HTTP errors
Non-2xx responses previously yielded bare `HTTP 403`. Now surface the
request URL, GitHub's JSON error message (or body snippet), X-RateLimit-Reset
when quota is exhausted, and Retry-After. Turns a mystery 403 into
'rate limit exhausted; resets at 2026-04-15T18:00:00Z' — the difference
between 'try GITHUB_TOKEN' and a wild goose chase.
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Skills don't nest. Once the manifest generator finds a valid SKILL.md
in a directory, it should not recurse into that skill's subdirectories
looking for more skills. Template files (like bmb's setup-skill-template)
inside a skill's assets/ would be incorrectly scanned and produce
spurious errors.
Add missing move() with cross-device fallback (rename → copy+rm on
EXDEV), needed by OfficialModules.createModuleDirectories for directory
migrations during upgrades.
Honor overwrite/errorOnExist options in copy() to match fs-extra
behavior for callers that pass these flags.
Dicky Moore
Dicky Moore
Brian
Jérôme Revillard
LanyGuan
Dicky Moore
AJ Côté
Curtis Ide
github-actions[bot]
Yahya Bin Naveed
Pablo Ontiveros
Murat K Ozcan
miendinh
Alex Verkhovsky
梁山河