claudecodex

build-loop

Keep everything you build honest. Log what a skill, hook, command, plugin or script got wrong the moment it happens, fix it from the queue with an approval gate, know what else the fix puts at risk, and keep a to-build list that closes itself when the work actually ships.

0 on GitHubby Sarah Maddenv0.10.24MITupdated 3 days agoproductivity
/

Plugin build-loop from marketplace infra-plugins in sarahcallmesmadds/infra-plugins. Commands read from the repo's manifest — 3 days ago.

This plugin runs code on your machine — its hooks fire automatically on events like tool use. The exact commands are listed below; read them before installing.

What it installs

Skills (11)

  • apply-fix

    Applies a correction from the bug queue to an actual target file. Reads the queue entry, checks DEPS.json for dependents, reasons about the surgical fix, shows a plain-language before/after diff, waits for the user's approval (yes / no / retry), then writes the fix, commits to the correct repo, and

  • audit-deps

    Scans every configured root (skills, hooks, commands), compares what is on disk to DEPS.json at ~/.claude/build-loop/, and surfaces missing (on disk but not in the map), orphaned (in the map but not on disk), or changed (the file is newer than its DEPS.json entry's last_updated) entries. Use when yo

  • built-check

    Cross-checks the to-build list at ~/.claude/build-loop/to-build/ against what has actually been built, and offers to close the finished items in one step. Looks for evidence in the git log of every configured root, on disk, and in the current session. Use at the end of a session, when wrapping up, o

  • devin-review-response

    Resolve a complete Devin code-review round without point fixes. Use when Devin posts PR findings, the user asks to address or fix a Devin review, or a branch needs a final Devin-response pass before push or merge. Captures every GitHub-app review and known CLI run for one commit, reconciles their re

  • find-skill

    Skill discovery — tells you which skill to use for what you're trying to do. Invoke when you're not sure which command to reach for. Builds the routing table at runtime by reading every SKILL.md frontmatter, so newly-installed skills auto-appear.

  • flag-issue

    Logs a correction to the bug queue at ~/.claude/build-loop/queue/, against anything the user built — a skill, a hook, a slash command, a plugin, or a loose script. Use when the user says "that was wrong", "it should have", "next time", "don't do that", corrects anything of theirs by name, says the o

  • flag-patterns

    A weekly report on what broke, what got fixed, and what keeps coming back. Reads the bug queue, finds the things corrected three or more times on three or more separate occasions, writes the report to ~/.claude/build-loop/summaries/YYYY-WW.md, and offers to post it to Slack. Use when the user asks "

  • list-bugs

    Shows the build loop bug queue as a plain-language table. Use when the user asks "what's in the queue", "what bugs are open", "show me the queue", "what did I capture", or explicitly invokes /list-bugs. Supports optional status filter argument (open, all, resolved, wontfix). Default filter is Open +

  • revert-fix

    Rolls back a committed fix. Reads the queue entry to find the commit hash, runs git revert in the correct repo (creates a new undo commit — does not delete or modify the original commit), resets the queue entry status back to Open, and stores the revert commit hash in notes. Works whether the commit

  • to-build

    The to-build list, at ~/.claude/build-loop/to-build/. With an argument it writes down something the user plans to build (a skill, hook, command, plugin, or loose script), showing a draft and waiting for confirmation before writing. With no argument it shows the list. Use when the user says "I want t

  • verify-fix

    Human-review verification gate for fixes. Presents the original failing scenario from the queue entry, shows the before/after diff, and asks the user whether the fix looks right. If yes — signals pass (apply-fix handles the commit). If no — leaves the queue entry Open with a note recording the rejec

Hooks (4)

  • PostToolUse

    [ -n "${CLAUDE_PLUGIN_ROOT}" ] || { printf '%s\n' "Plugin hooks are off because this host did not tell the hook where the plugin is installed. A restart will not help." >&2; exit 3; }; [ -e "${CLAUDE_

  • PostToolUse

    [ -n "${CLAUDE_PLUGIN_ROOT}" ] || { printf '%s\n' "Plugin hooks are off because this host did not tell the hook where the plugin is installed. A restart will not help." >&2; exit 3; }; [ -e "${CLAUDE_

  • UserPromptSubmit

    say(){ if [ -n "${PLUGIN_ROOT}" ] && [ "${PLUGIN_ROOT}" = "${CLAUDE_PLUGIN_ROOT}" ]; then printf '{"hookSpecificOutput":{"hookEventName":"UserPromptSubmit","additionalContext":"%s"}}' "${2:-$1}"; exit

  • UserPromptSubmit

    say(){ if [ -n "${PLUGIN_ROOT}" ] && [ "${PLUGIN_ROOT}" = "${CLAUDE_PLUGIN_ROOT}" ]; then printf '{"hookSpecificOutput":{"hookEventName":"UserPromptSubmit","additionalContext":"%s"}}' "${2:-$1}"; exit

View source on GitHub

More from sarahcallmesmadds/infra-plugins

All 6 plugins →

session

claudecodex

Make sessions continuous. Wrap one up into a handoff the next one loads, start with today's date and a compact build-loop brief, warn when another session is already working here, and put cost, context and connected tool health in the status line.

5 skills · hooks

0

git-hygiene

claudecodex

Tells you which branches are proved safe to delete, keeps project worktrees in one configured folder, and cleans up finished worktrees. It removes only the branches and worktrees you approve, re-checks each one first, and keeps anything it cannot prove safe.

4 skills · hooks

0

slop-check

claudecodex

Catch the habits that mean nobody edited it, and rewrite an answer that did not land. Blocks em dashes and choppy sentence runs in the assistant's own writing in both Claude Code and Codex, checks any draft, document, pull request, chart or spec for the signs it shipped unreviewed, reshapes a confusing answer around what you have to do, and makes each slide title state the takeaway of its slide.

3 skills · hooks

0

guardrails

claudecodex

Blocks commits to protected branches, asks before irreversible deletes, flags prompt injection in content the model reads or writes, and can require a governing document be read before a directory is written. Enforced automatically in both Claude Code and Codex.

2 skills · hooks

0

spend-guardrails

claudecodex

Choose the lowest-cost current Claude or OpenAI model that can reliably do the work, without recommending retired models.

1 skill

0

Similar plugins

Matt Pocock's agent skills for real engineering (grilling, spec/ticket flows, TDD, code review, domain modelling and more).

39 skills

mattpocock/skills 323,108

superpowers

claudecodex

Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques

15 skills · hooks

obra/superpowers 316,836

nextjs

claude

Official Next.js skills: adopt and optimize Cache Components, adopt Partial Prefetching, and verify runtime behavior against a running dev server.

6 skills

vercel/next.js 143,495

caveman

claude

Talk like caveman. Cut filler. Keep technical details explicit.

22 skills · 1 command · 3 agents · hooks

JuliusBrussee/caveman 119,462

claude-mem

claudecodex

Persistent memory for Claude Code: the next session starts where the last one left off.

29 skills · 1 MCP server · hooks

thedotmack/claude-mem 116,618

Taste Skill

claudecodex

tasteskill 2.5.1: professional company websites at the figma.com and x.ai level, with one section per screen, real photographs, real product UI, effects kept for tech and gaming pages, and a real-browser checker and finish gate.

1 skill

Leonxlnx/taste-skill 114,391

Maintain build-loop?

This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.

claude & codex plugins6 plugins