Claude-skill-registry feature-dev-loop

Execute feature lifecycle from beads issue to GitHub push without git locks or JSONL corruption. Triggers on "start feature", "implement issue", "feature workflow", "new feature".

install

source · Clone the upstream repo

git clone https://github.com/majiayu000/claude-skill-registry

Claude Code · Install into ~/.claude/skills/

T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/feature-dev-loop" ~/.claude/skills/majiayu000-claude-skill-registry-feature-dev-loop && rm -rf "$T"

manifest: skills/data/feature-dev-loop/SKILL.md

source content

Feature Dev Loop

Execute a complete feature lifecycle: issue → code → sync.

Prerequisites

```
bd
```
CLI available
Git repository initialized
```
.beads/
```
directory exists
No
```
.git/index.lock
```
file

Run

scripts/validate.sh

to verify.

Input Schema

Parameter	Type	Required	Default	Description
title	string	yes	-	Feature title for beads issue
type	feature\|bugfix\|chore	no	feature	Issue type
priority	P0\|P1\|P2\|P3\|P4	no	P2	Issue priority
scope	string	no	core	Commit scope
use_worktree	boolean	no	false	Create isolated worktree

Procedure

Step 1: Create Issue

@check: No existing issue with same title

@command:

ISSUE_ID=$(bd create --title="<title>" --type=<type> --priority=<priority> --json 2>/dev/null | grep -o '"id":"[^"]*"' | cut -d'"' -f4)
echo "Created: $ISSUE_ID"

@output: Issue ID (e.g.,

superpowers-abc

)

@on-error: Check

bd list

for duplicate titles

Step 2: Worktree Isolation (Optional)

@check:

use_worktree

is true AND branch does not exist

@command:

sp wt create feature/<slug> ../sp-<slug> --base=main
cd $(sp wt switch feature/<slug>)

@output: Working directory is isolated worktree

@on-error: If branch exists, use existing worktree or choose different name

Step 3: Mark In Progress

@check: Issue exists and is open

@command:

bd update $ISSUE_ID --status=in_progress

@output: Issue status changed to

in_progress

Step 4: Implement

@check: Issue is in_progress

Execute coding task:

Write code changes
Add tests if applicable

@command:

pnpm --filter @sp/<package> build
pnpm --filter @sp/<package> test

@output: Build succeeds, tests pass

@on-error: Fix errors before proceeding

Step 5: Finish (Atomic)

@check: Implementation complete, tests pass

Use the finish script for atomic close → sync → commit → push:

@command:

npx tsx .claude/skills/feature-dev-loop/scripts/finish.ts $ISSUE_ID "<type>(<scope>): <title>"

@output: Issue closed, JSONL synced, commit pushed

@on-error: Script provides recovery hints for each failure mode

The finish script (TypeScript):

Pre-flight checks: Validates git repo, bd CLI, issue exists, issue status
Closes issue with commit message as reason
Syncs beads (full sync, not just flush)
Commits with
```
BD_SKIP_AUTO_CREATE=1
```
to prevent auto-issue loop
Handles "nothing to commit" gracefully
Pushes with retry (auto-rebase on rejection, up to 3 attempts)
Provides recovery hints on failure with step-specific guidance

Success Criteria

```
bd show $ISSUE_ID
```
returns
```
status: closed
```
```
.beads/issues.jsonl
```
contains closed issue
```
git log -1
```
shows commit with issue ID
```
git status
```
shows clean working directory
Remote branch updated (check with
```
git fetch && git status
```
)

Error Recovery

The finish.ts script handles most errors automatically. Manual recovery only needed for:

Error	Recovery Action
Rebase conflicts	Resolve conflicts, `git rebase --continue && git push`
Branch exists	Use `--force` or choose different worktree name
Tests fail	Fix code, re-run tests before Step 5
Issue already closed	Manual commit: `git add . && git commit -m "..."`

Auto-handled by finish.ts:

```
index.lock
```
cleanup
Push rejection (auto-rebase retry)
Nothing to commit (graceful skip)
JSONL sync issues

Output Schema

{
  "status": "complete",
  "issue_id": "<beads-issue-id>",
  "commit": "<git-commit-hash>",
  "branch": "<branch-name>"
}