* test(bazel): add internal/testutil/bazeltest runfiles helpers
Repository-path resolution for tests that works identically under plain
go test (package cwd, full checkout), bazel test locally, and remote
execution (declared runfiles only):
- IsBazel, RepoRoot (BEADS_TEST_REPO_ROOT, then the runfiles workspace
root, then a go.mod walk), OverrideRoot (drop-in condition for existing
helpers; empty under plain go test), CallerDir
- Runfile for $(rlocationpath) values, GOROOT for type-checking tests
- ShardFreeEnv for helpers that re-exec a test binary
Every helper reduces to the pre-Bazel behavior under go test.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* test(bazel): declare source/data reads for 14 scan-guard targets
Source-scanning and repo-file tests read files their Bazel targets never
declared, so they failed under `bazel test` (and could not run on remote
workers at all). Declare narrow inputs as `data` and resolve paths through
bazeltest where the test located the tree via runtime.Caller:
- tools/bazel/go_srcs.py (run by `make bazel-sync` after gazelle) keeps a
managed `go_srcs` filegroup in the packages whose Go source tests scan,
plus `//internal/storage:tree_go_srcs` for the storage-tree walk. Bazel
globs stop at package boundaries, so the aggregate needs one filegroup
per package; the script keeps it in sync and gazelle-stable.
- Hand-written narrow inputs: //internal/storage/schema:migrations,
//:docsync_files, //plugins/beads:plugin_files, and exports of
.gitattributes, .claude-plugin/marketplace.json, cmd/bd list.go and
ready.go, testutil/testdoltcommon.go, scripts README.md and the release
catalog, each visible only to its consumer.
- runtime.Caller-based roots in backend, backend/conformance,
internal/storage, internal/formula and test/docsync now go through
bazeltest.CallerDir / OverrideRoot; relative-path tests need data only.
- integrations/ leaves .bazelignore (it is still gazelle-excluded) so the
docsync link check can declare integrations/beads-mcp/PYPI.md.
Under --config=prcore every target reports the same PASS/SKIP/FAIL test
set as `go test -race -short` for its package, locally and remotely.
Plain `go test` behavior is unchanged.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* test(bazel): guard generated go_srcs and harden bazeltest root override
- tools/bazel/go_srcs.py: add --check (print diff, exit 1 on stale blocks);
new `make bazel-sync-check` target runs it.
- scripts/bazel_policy_test.go: under go test, assert every tree_go_srcs
(//internal/storage) lists exactly the BUILD.bazel packages below its
root, run go_srcs.py --check, and forbid BUILD files under docs/ and
engdocs/ (they would cut subtrees out of //:docsync_files).
- bazeltest: honor BEADS_TEST_REPO_ROOT only when it holds go.mod
(RepoRoot fails loudly otherwise); drop the unused GOROOT() helper,
whose compiled-in branch could never fire under rules_go.
- scripts/ci/lib/test-env.sh: unset BEADS_TEST_REPO_ROOT.
- Document the hand-listed out-of-tree link targets in //:docsync_files.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* build(bazel): re-sync BUILD files with main
Regenerated with make bazel-sync after commits that landed while the
scaffold PR was in CI (new internal/execenv, gitenv, gitignore packages
and new sources/deps).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* test(bazel): declare init_safety_help.go for docsync
TestInitSafetyCLIHelpCoversReCloneGotchas (#6431) reads
cmd/bd/init_safety_help.go; export it to //test/docsync only.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* build(bazel): add side-by-side Bazel scaffold (module + gazelle)
Add a Bzlmod Bazel build alongside the Go toolchain. `go build` and
`go test` are unchanged; nothing in CI invokes Bazel yet.
- MODULE.bazel: rules_go 0.63.0, gazelle 0.53.0, Go SDK 1.26.7 (matches
the go.mod toolchain directive), go_deps from go.mod.
- ICU policy (engdocs/ICU-POLICY.md): a go-mysql-server gazelle_override
selects the pure-Go regex backend and excludes regex_cgo.go, and
.bazelrc builds with --@rules_go//go/config:tags=gms_pure_go, so no
target depends on go-icu-regex or links libicu.
- third_party/patches: three module_override patches for dolt deps that
ship WORKSPACE-era BUILD files, adapted from the gascity project.
- .bazelrc: prcore/ci test configs (inert for now; they match pr-core's
test selection, not yet its environment isolation) and a generic
remote-exec config with a bounded remote timeout. Executor endpoints and
TLS credentials come only from a gitignored .bazelrc.local/user.bazelrc
or a CI rc outside the workspace.
- .bazelversion pins Bazel 9.2.0; .bazelignore skips separate modules,
non-Go trees, nested agent worktrees (.claude, .worktrees) and local
data/output dirs (.beads, .dolt, dist, .runtime, tmp), so a primary
checkout neither builds nor gazelle-rewrites nested worktree copies.
- Gazelle-generated BUILD.bazel files for every Go package.
- Makefile: `make bazel-sync` (gazelle + bazel mod tidy).
- .gitignore: bazel-* output symlinks and machine-local rc files.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* test(bazel): policy tests for the Bazel configuration
Add scripts/bazel_policy_test.go, plain Go tests that run under both
`go test ./scripts` and `bazel test //scripts:scripts_test`:
- MODULE.bazel go_sdk.download version equals the go.mod toolchain.
- ICU policy: .bazelrc sets gms_pure_go for build, the go-mysql-server
gazelle_override excludes regex_cgo.go, and the root BUILD.bazel
carries `# gazelle:build_tags gms_pure_go`.
- .bazelversion pins an exact Bazel release.
- .bazelrc.local and user.bazelrc are gitignored (parsed from .gitignore
and cross-checked with git check-ignore when in a checkout).
- No remote-execution endpoints or credentials in committed bazelrc-like
files, Markdown docs or .github/**: remote/BES/TLS flags
(--remote_executor, --remote_cache, --bes_backend, --remote_header,
--tls_client_certificate, ...) with a literal value, with or without a
URL scheme, and non-loopback gRPC URLs in rc/.github files. Placeholders,
$VARIABLES, and loopback are allowed; Markdown gRPC URLs (OTel exporter
docs) are not flagged. Skipped outside a git checkout, e.g. in the Bazel
sandbox.
Each check is a pure function also exercised against fixtures that break
the invariant. The files the tests read are declared as data via
//:bazel_policy_files so they resolve from runfiles under Bazel.
scripts/check-build-tags.sh now treats Bazel invocations as governed by
.bazelrc: the shell command segment that invokes Bazel (split on &&, ||,
;, |, $( and backticks) is exempt from the per-line `go` tag check, while
a bare `go` in another segment of the same line is still flagged. When
any scanned file invokes Bazel (or .bazelrc exists) .bazelrc must set
--@rules_go//go/config:tags=gms_pure_go for build.
scripts/check_build_tags_test.go covers both in a temp repository.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(scripts): split bazel lines in pure bash so check-build-tags works with BSD sed
BSD sed (macOS) does not expand \n in s/// replacements, so the segment
split never happened and the Bazel-exemption tests failed on macos-latest.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* test(scripts): skip check-build-tags tests when bash lacks mapfile
macOS runners use /bin/bash 3.2, which has no mapfile; check-build-tags.sh
has always required bash >= 4, so the new tests are skipped there.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(setup/claude): keep settings.json's trailing newline and skip no-op rewrites (#5693)
bd setup claude (and bd init, which calls it) marshalled the Claude settings
map with json.MarshalIndent and wrote it unconditionally. MarshalIndent emits
no trailing newline, so every run stripped the one Claude Code's own writes
leave behind, and re-wrote the file even when no hook changed. Because bd init
git-adds .claude/settings.json, that surfaced as a staged diff with no
semantic content.
The same marshal-then-write pattern appears at four call sites in claude.go:
installClaude's settings write and its settings.local.json migration, and
removeClaude's settings write and its settings.local.json cleanup. Fixing one
would have left the bug live in the other three, so both halves are now shared
helpers -- marshalSettings appends the newline, writeSettingsIfChanged compares
against the bytes already on disk and returns early when they match -- and all
four sites go through them.
Reported by @maphew.
* chore(setup/claude): use the repo spelling "marshaled" in the new comments
The misspell linter flags British "marshalled"; CI pins golangci-lint
v2.10.1 and runs it with --new-from-merge-base, so both new comment lines
are in scope. Comment text only, no behaviour change.
#4838 taught `bd setup claude` to detect a CLAUDE.md that is a thin stub
importing AGENTS.md, and to write the managed beads block to AGENTS.md instead.
isAgentsImportStub decides that by scanning for "@AGENTS.md" on its own line,
with no notion of code blocks — so a CLAUDE.md that merely *documents* the
pattern in a fenced example is classified as a stub.
That is not a cosmetic misread. Once the redirect activates, stripStaleClaudeBlock
removes the managed block from CLAUDE.md, on the reasoning that AGENTS.md now
owns it. For a file that was in fact authoritative, a documentation example
silently moves the block out of it. The new end-to-end test is the one that
matters here: against the parent commit it fails with
CLAUDE.md is authoritative and must carry the beads block;
a fenced example redirected it away
Skipping fenced regions also matches the semantics being detected: Claude Code
does not expand an @-import shown as code, so a fenced directive was never going
to make the file a stub in the first place.
Deliberately narrow in two places, both pinned by tests:
- Four-space-indented lines are still scanned. Indentation is ambiguous in a way
a fence is not — it is equally a list continuation, which is a plausible home
for a real directive — so treating it as code would trade this false positive
for a false negative. "indented directive still counts" pins that.
- A closing fence must match the character that opened the block, so ``` inside a
~~~ block does not terminate it and let a following directive leak through.
An unterminated fence swallows the rest of the file, which is also what a
Markdown renderer does, so the conservative answer there is the correct one
rather than an accident. Both are pinned.
Five of the seven new unit cases fail against the parent commit; the two that
pass in both directions are the over-correction guards ("real directive after a
fenced example still counts", "indented directive still counts") and exist so a
future fix cannot go too far the other way.
This closes the last of the two nits left open on #4838's review; the other
(path-qualified "@./AGENTS.md") was already handled before that PR merged.
Note for whoever looks at CI: `main` is currently red on Test (macos-latest) for
three unrelated failures — see #5210 and #5220. This branch touches only
cmd/bd/setup and does not go near them.
Agent-Signature: claude-opus-5-high on behalf of maphew
Normalize every shared command argument-hint to an explicit YAML string and enforce the schema across LF and CRLF checkouts so Copilot plugin skills load reliably.
Agent-Signature: codex-gpt-5.6-sol-high on behalf of matt wilkie
* feat(setup): redirect beads block to AGENTS.md when CLAUDE.md is a stub
When CLAUDE.md is a thin pointer that imports AGENTS.md via the
@-include convention (e.g. "@AGENTS.md" on its own line), bd setup claude
now writes the managed BEADS INTEGRATION block to AGENTS.md instead of
duplicating it in the stub. This supports the shared-authoritative-file
pattern used by repos that keep AGENTS.md as the single source of agent
instructions.
The redirect applies to install, check, and remove operations. If
AGENTS.md does not exist, the fallback behavior (writing to CLAUDE.md)
is preserved.
Agent-Signature: amp-gpt-5.4-2026-03-05-high on behalf of maphew
* fix(setup): strip stale CLAUDE.md beads block when AGENTS.md redirect activates
An older bd could have written a beads-managed block directly into CLAUDE.md
before a project adopted the "@AGENTS.md" import-stub pattern; once the
redirect to AGENTS.md activates, that block became an orphaned duplicate that
neither install nor remove ever cleaned up. installClaude now strips a stale
block from CLAUDE.md when the redirect is active (preserving the rest of the
stub), and removeClaude strips it too, alongside removing the AGENTS.md copy
as before.
Agent-Signature: claude-fable-5-high on behalf of matt wilkie
* fix(setup): gate CLAUDE.md strip on successful AGENTS.md injection; preserve shared AGENTS.md block on remove
Addresses two major findings from PR #4838 review:
- installClaude stripped the stale CLAUDE.md beads block whenever the
AGENTS.md redirect was active, even when installAgents skipped injection
(e.g. AGENTS.md is a symlink). That deleted the only copy of the beads
block without writing a replacement. Gate the strip on
`redirected && !agentsSkipped`.
- removeClaude with the redirect active called removeAgents against
AGENTS.md, deleting the project-authoritative shared beads section that
bd init/other agents' setup created and other agents rely on. Limit the
redirected remove path to stale CLAUDE.md cleanup and leave the shared
AGENTS.md section intact.
Also addresses minor finding: isAgentsImportStub now also accepts the
`@./AGENTS.md` relative-import form, not just `@AGENTS.md`.
Deferred (minor, ambiguous product intent, needs maintainer input rather
than a guess): blank AGENTS.md gets ProfileMinimal on redirect; whether it
should get the full profile instead is a product decision, not a bug fix.
Added regression tests: TestInstallClaudeRedirectSkipsStripWhenAgentsSymlink,
TestRemoveClaudeRedirectPreservesFullProfileSharedBlock; updated
TestRemoveClaudeRedirectsToAgentsMD and
TestRemoveClaudeRedirectCleansStaleClaudeBlock to assert the shared AGENTS.md
block is preserved on remove; added an @./AGENTS.md case to
TestIsAgentsImportStub.
Agent-Signature: claude-fable-5-high on behalf of matt wilkie
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Recognize the exact historical bd-prime SessionStart pipeline, remove only that nested command, preserve unrelated hooks, and report the legacy-plus-canonical state as stale.
Agent-Signature: codex-gpt-5.6-sol-high on behalf of matt wilkie
Per decision 7, bd's user-facing strings now reference the Mintlify-era
paths: prime policy-profiles help, storage-backend comparison help
(store_factory_nocgo, init --backend), init-safety help (recovery
playbooks now docs/recovery/init-safety.md; the ADR reference now
engdocs/adr/), the pk-fork-refused recovery URL (dolt.go), doctor's plugin
suggestions (claude.go) and managed-handoff pointer, aider setup guidance
and the shared recipe template (quickstart URL), and the agent instruction
templates (sync-concepts URL, quickstart URL). bd mol pour's help no
longer claims molecules are 'synced with git' — they sync via Dolt like
any other bead.
Also corrects the stale claude recipe ProjectPath metadata
(.claude/settings.local.json -> .claude/settings.json; the dedicated
installer has written settings.json since the migration sweep).
This repo's AGENTS.md managed section is refreshed to the new template
body (hash bacef91e, has-remote variant preserved), and agents.md.tmpl's
embedded marker matches. Template tests updated to pin the new fragment;
CLI docs regenerated.
Recorded, not fixed (behavior-level, see the decision record):
routing.mode's maintainer/contributor no-op values and the
sync.require_confirmation_on_mass_delete key with no consumer.
Note: TestCheckBeadsRole_NotConfigured/_NotGitRepo in cmd/bd/doctor fail
on any machine with beads.role in the global git config (no HOME
isolation) — pre-existing, unrelated to this change.
Cursor support is anemic compared to Claude Code and Codex. It ships
only a static rules file: no agent lifecycle hooks, so `bd prime`
context is never injected at session start and is silently lost after
context compaction, and it has no `bd init` auto-install, no `--global`
scope, no shared agent skill, and no `bd doctor` coverage.
This PR brings Cursor up to the same footing as the tier-1 integrations.
- `bd setup cursor` now installs agent lifecycle hooks
(`.cursor/hooks.json`) alongside the rules file, wiring
sessionStart/preCompact/postToolUse to a hidden `bd cursor-hook`
command so Beads context survives compaction:
* sessionStart injects full `bd prime` context into new sessions
* preCompact arms a one-shot refresh marker
* postToolUse re-injects `bd prime` once after a compaction, then
no-ops
- `bd init` auto-installs Cursor (rules + skill + hooks) and stages
`.cursor/`
- `bd setup cursor --global` writes `~/.cursor/hooks.json` +
`~/.agents/skills/beads`
- Canonical rules: `.cursor/rules/beads.mdc` wraps the shared
`recipes.Template` instead of a hand-maintained copy
- Shared agent skill (`.agents/skills/beads`), removed only once neither
Cursor nor Codex still relies on it
- `bd doctor` reports Cursor Integration, Settings Health, and Hook
Completeness checks with agent-mode enrichment
Shared hook mechanics (prime runner + one-shot refresh marker) are
factored into `cmd/bd/agent_hook.go` and reused by codex-hook.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat: Add bd unclaim command to release stuck issues
- Add UnclaimIssue method to storage interface and implementations
- Add bd unclaim CLI command: clears assignee, resets status to 'open'
- Improve claim error message when issue has assignee but status is 'open'
(now suggests using bd unclaim)
- Update documentation: CLI_REFERENCE.md, QUICKSTART.md, setup templates
- Add unclaim tests
- Fix telemetry wrapper to implement UnclaimIssue method
- Add test mocks for UnclaimIssue in jira tests
* docs: regenerate CLI reference for bd unclaim
Runs scripts/generate-cli-docs.sh to add the bd unclaim entry that
gastownhall/beads#3694 introduced but whose CLI_REFERENCE.md/website
delta was lost in the conflict resolution during the cherry-pick.
Scoped to only the unclaim addition; unrelated doc drift already
present on main (formula/lint/types/linear entries) is left untouched.
Agent-Signature: claude-fable-5-medium on behalf of matt wilkie
---------
Co-authored-by: Abner S. Alves Junior <abnersajr@gmail.com>
* feat(no-push): guard dolt push/pull and strip push from rendered templates
- bd dolt push/pull skip silently (exit 0) when no-push: true in config
- RenderOpts gains NoPush bool; !HasRemote || NoPush strips bd dolt push
- NoPush propagated through AgentsFileParams, fs_adapters, init, setup/agents, init_proxied_server
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(no-push): anchor stripDoltPushReferences to prevent 4-space indent false match
Without the newline anchor, ReplaceAll matched any line containing " bd
dolt push\n" as a substring — including 4-space indented code blocks that
happen to share the prefix. Pad body with newlines so the match is anchored
to the start of the line.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(no-push): remove pull guard from bd dolt pull (be-ve2x6)
no-push means "do not publish local state to the remote"; it must not
prevent receiving remote updates. The guard at dolt.go:330 made
bd dolt pull return success without contacting the remote when
no-push: true, leaving contributor clones stale.
Remove the pull guard. Add TestNoPushDoesNotSkipDoltPull to prevent
reintroduction: injects a minimal fake store and asserts Pull() is
called even under no-push: true.
Push guard (dolt.go:257) and template stripping (render.go:217)
are unchanged — both are correct behavior per maphew's review of #4212.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* test(dolt): add TestNoPushSkipsDoltPush for push guard coverage (be-i9e2)
Direct unit test for the no-push guard at cmd/bd/dolt.go:247-249.
Asserts Push() is not called and output contains "skipping push"
when BD_NO_PUSH=true. Mirrors TestNoPushDoesNotSkipDoltPull pattern.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* chore: release gate PASS for be-i9e2 (no-push dolt push test)
* test(dolt): fix minimalPullStore for current main (ListRemotes drift)
Rebasing #4323 onto current main surfaced a latent defect hidden by cached
green checks: main now routes bd dolt pull failure handling through
isConfirmedNoRemote, which calls st.ListRemotes during error classification
(cmd/bd/dolt.go). minimalPullStore embeds a nil storage.DoltStorage and
overrode only Pull, so TestNoPushDoesNotSkipDoltPull nil-panicked once merged
into current main.
Add a ListRemotes that reports no remotes, keeping the simulated pull failure
on the benign no-remote path (clean exit) so the test still proves the no-push
guard does not short-circuit bd dolt pull. Same fix quad341 applied on #4212.
Verified against current main: TestNoPushSkipsDoltPush, TestNoPushDoesNotSkipDoltPull,
TestIsConfirmedNoRemote, TestDoltPushPullCommitNeedStore all pass.
Agent-Signature: claude-opus-4-8-high on behalf of maphew
---------
Co-authored-by: Jim Wordelman <jim@wordelman.name>
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: matt wilkie <maphew@gmail.com>
* fix(setup): match beads plugin name exactly, not by substring (#4244)
bd setup claude decided whether the beads hook plugin was enabled with a
substring test, strings.Contains(lower(key), "beads"). Any enabledPlugins
key merely containing "beads" (e.g. design-to-beads@xexr-marketplace) was
mistaken for the beads hook plugin, so the SessionStart hook write was
skipped and the hook landed nowhere.
enabledPlugins keys are "<pluginName>@<marketplace>"; match the plugin-name
segment exactly instead. A real beads@<marketplace> still matches, so the
GH#3192 double-fire prevention is preserved.
- Split the key on the first "@" and compare the name == "beads"
- Add regression subtests: design-to-beads alone is not detected; a real
beads plugin is still found alongside a *beads* decoy
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(setup): match codex beads plugin name exactly, not by substring (#4244)
bd setup codex decided whether the beads hook plugin was enabled with a
substring test, strings.Contains(lower(table), "beads") on the TOML plugin
table header. Any [plugins.<name>] whose name merely contains "beads" (e.g.
design-to-beads) was mistaken for the beads hook plugin, so the hooks
fallback write was skipped — the codex sibling of the claude.go bug already
fixed in this PR.
Codex plugin tables are [plugins.<name>] or [plugins."<name>@<marketplace>"]
(TOML quotes a key containing "@"). Strip the "plugins." prefix and any
quotes, then match the <name> segment of <name>@<marketplace> exactly,
mirroring the claude.go fix. A real beads@<marketplace> still matches, so the
GH#3192 double-fire prevention is preserved.
- Exact-match the plugin-name segment instead of a substring test
- Add TestCodexConfigEnablesBeadsPlugin: design-to-beads (quoted and bare) is
not detected; a real quoted/bare beads plugin still is, including alongside
a *beads* decoy; a disabled beads plugin is not detected
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
PR #4188 moved hook files from plugins/beads/hooks/ to
plugins/beads/.codex-plugin/hooks/ to stop Claude from discovering
Codex-only hooks. However, the hooks path in .codex-plugin/plugin.json
remained "./hooks/hooks.json" which Codex resolves relative to the
plugin root (plugins/beads/), not relative to .codex-plugin/. This
caused a runtime error:
failed to read .../beads/1.0.5/hooks/hooks.json: No such file or directory
Fix: update the hooks path to "./.codex-plugin/hooks/hooks.json" so
it correctly resolves to plugins/beads/.codex-plugin/hooks/hooks.json.
Update TestPluginLayoutUsesSharedBeadsRoot to assert the corrected path.
Fixes#4210. Follow-up to #4188.
Refuse setup atomic writes through symlinks unless a caller explicitly opts into following the link, and report partial Claude/Mux setup when agent instructions are skipped.
Document repair steps for repositories already affected by corrupted mode-120000 CLAUDE.md entries.
Refs #3722.
_codex-gpt-5.5-medium on behalf of matt wilkie_
Move Codex native hook metadata under the Codex-only plugin directory so Claude keeps using only its inline `bd prime` hooks.
Leave release version bumps to the maintainer release flow while documenting hook metadata isolation under Unreleased.
Refs #3924.
_codex-gpt-5.5-medium on behalf of matt wilkie_
Warn on stderr with the symlink target and a concrete recovery path when setup skips managed-section injection through an agent-file symlink. This keeps the corruption guard explicit without pretending the integration text was written.
Also cover the Lstat inspect-error branch called out in PR review coverage.
_codex-gpt-5.5-medium on behalf of matt wilkie_
Avoid following symlinked agent instruction files when installing managed setup sections. This preserves symlink mode and target content while still allowing normal files to receive integration text.\n\nRefs gastownhall/beads#3722\n\n_codex-gpt-5.5-medium on behalf of matt wilkie_
Run Codex project setup during `bd init` using the existing setup installer, and stage generated Codex project files in the bootstrap commit.
Also refresh the generated setup/init docs for the new default behavior.
Co-authored-by: Eugene Brevdo <ebrevdo@gmail.com>
_codex-gpt-5.5-medium on behalf of CI Bot_
* fix(init): wire git origin as Dolt remote
Default bd init now treats a git origin as the Dolt remote, persists sync.remote, and still skips synthesis in --stealth. JSONL auto-export/import paths now warn or skip based on sync.remote so exports do not masquerade as cross-machine sync.
Docs and generated agent/CLI references now state that Dolt is canonical, refs/dolt/data is the sync channel, and .beads/issues.jsonl is export/interchange/backup only.
_codex-gpt-5-xhigh on behalf of maphew_
* test: keep embedded command stderr out of parsed stdout
Embedded command tests parse IDs and JSON from command output. The new auto-export warnings correctly go to stderr, but several tests used CombinedOutput and treated warnings as stdout, which broke CI shards that parse issue IDs or JSON lists. Capture stdout separately on success while preserving combined output for error diagnostics.\n\n_codex-gpt-5-medium on behalf of CI Bot_
---------
Co-authored-by: CI Bot <ci@beads.test>
Fixes Gemini CLI hook integration by emitting the required SessionStart JSON envelope. Extends the same standard to Claude Code hooks (documented spec, not just tolerated raw text). Adds legacy detection in bd setup gemini --check.
- Add `--hook-json` flag to `bd prime` wrapping output in the cross-CLI `hookSpecificOutput.additionalContext` JSON envelope (Claude Code, Gemini CLI, and Codex all share this contract)
- Update `bd setup gemini` to register `bd prime --hook-json` on SessionStart; remove PreCompress registration (advisory-only, no additionalContext support per Gemini docs v0.40.1)
- Update `bd setup claude` to register `bd prime --hook-json` on SessionStart and PreCompact (JSON form is the documented spec; raw text was merely tolerated)
- Migration sweeps in both setup commands so re-running upgrades bare-command installs cleanly
- `bd setup gemini --check` distinguishes current (`--hook-json`) from legacy (bare `bd prime`) installs and emits upgrade advisory for legacy
- Fix `removeHookCommand` bug: previously dropped entire hook entry on first command match, silently deleting sibling commands; now filters only the matching command and drops the entry only when empty
- Refresh docs and recipes to reflect SessionStart-only Gemini hook behavior
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Test hardcoded forward-slash path literals against output from
filepath.Join, so it failed on Windows even though the production
code (geminiProjectSettingsPath, geminiGlobalSettingsPath) was
correct.
Matches the existing pattern from TestMuxProjectAgentsPath and
TestMuxGlobalAgentsPath in the same package, which compare
filepath.Join(...) on both sides.
Fixes#3554
Mux post-edit hook now calls bd dolt push when only ~/bin/bd is available,
matching the primary branch.
bd setup claude --check scans Claude settings for removed bd sync hook
commands and prints a stderr warning with remediation.
Tests: TestClaudeSettingsUsesRemovedSyncCommand.
When removeHookCommand found a matching command it set keepHook = false and broke out of the loop, discarding the entire hook entry rather than just the matching command. Any sibling commands in the same entry were silently deleted.
Fix: build a remaining slice by filtering only the matching command. The hook entry is dropped only when remaining becomes empty, preserving all sibling commands.
Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>
the claude plugin manifest declared agents at ./skills/beads/agents/, which
also contains openai.yaml (codex interface metadata). claude code's plugin
loader scans that dir and rejects non-md files, failing manifest validation
with 'agents: Invalid input'.
move task-agent.md back to plugins/beads/agents/ (the claude default, what
v1.0.0 used) and drop the explicit agents key from .claude-plugin/plugin.json.
codex's openai.yaml stays at skills/beads/agents/openai.yaml where bd setup
codex writes it and where the embed in internal/templates/ lives.
shared skill tree and dual claude/codex manifests are preserved per #3664.
Fixes#3741
RenderSection() and ReplaceSection() now accept RenderOpts with a
HasRemote flag. When false, "bd dolt push" is stripped from the
session-completion code block (and the Auto-Sync informational bullet
in the full profile). This prevents agents from dutifully running a
no-op every session on solo/local-only projects.
bd init detects the remote via shouldWireInitRemote(); bd setup detects
it by reading sync.remote / sync.git-remote from config.yaml. The
existing no-arg RenderSection() and ReplaceSection() default to
HasRemote=true for full backward compatibility.
Closes beads-ozg.
When both the beads Claude Code plugin and `bd setup claude` are active,
bd prime fires twice per session — wasting ~1-2k tokens each time. The
plugin already declares identical SessionStart/PreCompact hooks in its
plugin.json, so project-level hooks are redundant.
Now `bd setup claude` detects the plugin via enabledPlugins in settings
files and skips writing hooks when the plugin is already providing them.
The CLAUDE.md integration section is still installed since the plugin
does not provide that. `bd setup claude --check` reports "plugin-managed"
as a valid success state instead of "No hooks installed".
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Without YAML frontmatter, Cursor defaults .mdc files to Manual mode,
requiring users to @-reference the rule explicitly. Add alwaysApply: true
so beads context loads automatically every session, consistent with the
Claude and Gemini integrations.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Changes:
- bd setup claude now defaults to project-local (.claude/settings.json)
instead of global (~/.claude/settings.json)
- bd setup claude --global explicitly installs globally
- bd init automatically runs Claude hook setup (project-local)
- Legacy .claude/settings.local.json hooks are auto-migrated on install
and cleaned on remove
- bd setup claude --check detects legacy installs with migration hint
- bd init stages .claude/settings.json and CLAUDE.md for git commit
Closes#2935
Amp-Thread-ID: https://ampcode.com/threads/T-019d4fc1-6587-75cb-bdcb-1912a98cb1c1
Co-authored-by: Amp <amp@ampcode.com>
Add --agents-file flag to bd init allowing users to specify a custom
filename for agent instructions (default remains AGENTS.md). This
addresses the use case where projects prefer different naming
conventions (e.g. BEADS.md) without requiring a post-init rename.
Changes:
- Add config.AgentsFile() accessor and config.ValidateAgentsFile()
with security validation (path traversal, symlinks, extension check)
- Add --agents-file flag to bd init with config persistence
- Thread configured filename through init_agent.go
- Update setup/agents.go to use config.AgentsFile()
- Update doctor checks to search configured filename
- Update onboard messaging to reference configured filename
- Add DefaultAgentsFile constant for consistent fallback
The filename is persisted to .beads/config.yaml under agents.file
and read by all subsystems via config.AgentsFile().
Fixes#2737
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Introduce ErrServerNotRunning sentinel so callers can distinguish "already
stopped" from real failures. Stop now cleans up leftover PID/port files even
when the server is not running, preventing bd dolt status from reporting stale
state. Consolidate IsAutoStartDisabled into a single exported function covering
both env var and config sources.
PR: #2740
Co-Authored-By: Harry Miller <harry-miller-trimble@users.noreply.github.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Create agents.ReplaceSection(content, profile) as the single canonical
function for replacing beads integration sections. Both init_agent.go
(upgradeBeadsSection) and setup/agents.go (updateBeadsSectionWithProfile)
now delegate to it.
Key changes:
- Hash-based freshness check skips rewrite when content is current
- Returns error on malformed markers (missing/reversed) instead of
silent append
- Marker format now includes v:1 version field for forward compat
- atomicWriteFile accepts optional permissions param (default 0644)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Task 8: bd init now upgrades legacy beads sections to versioned format
instead of silently skipping them. New files created via EmbeddedDefault()
also get versioned markers.
Task 9: Replace stale bd sync references with bd dolt push in recipes
template. Add bd prime SSOT messaging.
Task 10: installAgents resolves symlinks via ResolveForWrite before
reading content. Profile precedence: if file already has full profile
and minimal is requested, full is preserved to avoid information loss.
Task 11: Added tests for staleness detection, profile extraction,
profile precedence, symlink safety, and legacy-to-new migration.
checkAgents now reports stale/current/missing status.
Task 12: Updated SETUP.md with prime SSOT, profiles, and profile
precedence documentation. Updated CLI_REFERENCE.md with --check
behavior and Gemini integration description.
Amp-Thread-ID: https://ampcode.com/threads/T-019ced7a-0c86-708d-930c-ee35d00eda8d
Co-authored-by: Amp <amp@ampcode.com>
When running 'bd setup claude --project' or 'bd setup gemini --project',
after installing hooks in settings.json, also install a minimal-profile
beads section in AGENTS.md. This provides a quick reference pointer to
'bd prime' for the full workflow, while hooks handle the heavy lifting.
On removal, the beads section is cleaned up from AGENTS.md as well.
The minimal profile contains: quick reference commands, pointer to
bd prime, and the landing-the-plane session close protocol — without
the full issue types, priorities, and sync details that hook-enabled
agents get dynamically from bd prime.
Amp-Thread-ID: https://ampcode.com/threads/T-019ced5e-0acd-7158-93ad-0695c6d02580
Co-authored-by: Amp <amp@ampcode.com>
Set profile: agents.ProfileFull explicitly on codex, factory, mux
(all three layers), and opencode integrations. This makes the profile
intent clear and prepares for adding minimal-profile integrations.
Amp-Thread-ID: https://ampcode.com/threads/T-019ced5e-0acd-7158-93ad-0695c6d02580
Co-authored-by: Amp <amp@ampcode.com>
Implement the foundational pieces for making bd prime the single source
of truth for agent instructions:
Task 1 - Shared render API (internal/templates/agents/render.go):
- Profile type with Full and Minimal constants
- RenderSection(profile) generates sections with marker metadata
- ParseMarker() extracts profile/hash from begin markers
- CurrentHash(profile) for staleness detection
- 8-char SHA-256 hash for deterministic freshness
Task 2 - Template splitting:
- New beads-section-minimal.md for hook-enabled agents (Claude, Gemini)
- Minimal profile: pointer to bd prime + quick reference + landing protocol
- Existing beads-section.md preserved as full profile body
Task 3 - Parser refactor (cmd/bd/setup/agents.go):
- Prefix-based marker matching via containsBeadsMarker/findBeginMarker
- Legacy markers (no metadata) detected and upgradable
- New markers: <!-- BEGIN BEADS INTEGRATION profile:full hash:a1b2c3d4 -->
- agentsIntegration gains profile field (defaults to full)
- updateBeadsSectionWithProfile for profile-aware updates
- createNewAgentsFileWithProfile for new file creation
Backward compatible: EmbeddedBeadsSection/EmbeddedDefault still work.
Legacy markers are auto-upgraded on next bd setup run.
Amp-Thread-ID: https://ampcode.com/threads/T-019ced5e-0acd-7158-93ad-0695c6d02580
Co-authored-by: Amp <amp@ampcode.com>
bd sync was removed as a command but references persisted in:
- CLI help text (--dolt-auto-commit flag description)
- Code comments (dolt_autocommit.go, init_contributor.go)
- Agent setup generators (aider, cursor, junie, mux)
- Doctor agent remediation commands
- Jujutsu hook alias (init_git_hooks.go)
- Copilot integration docs
All references now point to the current commands:
- bd dolt commit (for committing changes)
- bd dolt push (for pushing to remote)
Tests updated to match.
Complements #2389 (config/sync.mode removal) and #2419 (command removal).
Remaining stale references in other docs tracked in GH#2435.