repacted

Plugin repacted from marketplace repacted in v4lheru/repacted-plugin. Commands read from the repo's manifest — 3 days ago.

Skills (28)

• architecture-heuristics

  Use when designing or reviewing the shape of a system or change: component boundaries and C4 sketches, API and producer/consumer contracts, impact on other readers, reconcile and sync design, idempotency ledgers, data integrity, re-aggregate vs re-extract, anti-patterns to avoid.

• concurrent-agent-git

  Use before git add, commit, stash, checkout, reset, rebase or a build in any worktree or scratch dir other agents may share, and whenever a result depends on which tree, branch, index or session root you are in.

• frontend-verification

  Use when building or verifying UI (React, Next.js, Astro, Tailwind v4, CSS) or judging UI from a screenshot, DOM or Playwright run: build-only breaks, tokens, theming and fonts, layout and specificity, focus, motion and a11y, capture traps, proving the rendered state.

• google-workspace

  Use for Google Apps Script and clasp (push, deploy, re-pinning web-app deployments, .claspignore, triggers, quotas, LockService, time zones) and the Google Drive, Docs, Sheets and Gmail APIs (shared drives, export, batch updates).

• llm-pipeline-patterns

  Use when an LLM is a stage in a pipeline or product (Anthropic or OpenAI APIs, structured output, caching, classifiers, evals, map-reduce reports): field-observed provider traps, output validation and safety, extending a frozen report without a re-run. Check claude-api first for model IDs and pricin

• memory-hygiene

  Use when writing, correcting or migrating agent memory: MEMORY.md heads and the inbox, pact-memory records, correcting a claim in every carrier, the distill pass that moves inbox lessons into skills (distill-inbox), and index budgets (check-index).

• multi-agent-collaboration

  Use when you are a PACT teammate and another agent''s words drive your action: a dispatch premise that may be false, a ruling or relayed claim, instructions that contradict the code or each other, scope drift, whether to hold, ask or proceed, or how to report back to the lead.

• pact-architecture-patterns

  Use when designing in the ARCHITECT phase or reviewing architecture: C4 diagrams, component design and boundaries, API contracts and interfaces, design patterns, microservices vs monolith, anti-patterns.

• pact-coding-standards

  Use when writing, refactoring or reviewing code in the PACT Code phase: clean-code principles, error handling, logging, naming, and the lint and unused-import checks.

• pact-field-notes

  Use when PACT itself misbehaves: resume or compact team-name mismatch, dispatch_gate "no Task assigned", task not found or metadata missing, HANDOFF harvest gaps, worktree write-guard pins, hook ask/deny semantics, merge-guard approval format, pact-memory recall or --replace surprises.

• pact-handoff-harvest

  Secretary HANDOFF harvest, script-first: Standard, Incremental, Consolidation, orphans. Triggers: harvest HANDOFFs, process HANDOFFs, incremental, consolidation, handoff recovery.

• pact-lead-playbook

  Use when you are the PACT lead: writing a dispatch brief, stating premises, ruling or relaying mid-flight, accepting or rejecting a teachback or HANDOFF, deciding whether to delegate (by default investigations stay with the lead), sizing a HALT, calling the stop on review loops, launching long jobs,

• pact-memory

  PACT institutional memory (SQLite): save and search decisions, lessons and entities across sessions. Triggers: memory, save memory, search memory, lessons learned, remember, recall

• pact-prepare-research

  Use when researching in the PREPARE phase: technology and library evaluation, comparison matrices, API documentation, framework options, requirements gathering.

• pact-security-patterns

  Use when doing a security audit, auth flows, API credential handling, sensitive data (PII, financial, health) or vulnerability review: OWASP, encryption, backend proxy, frontend credential exposure.

• pact-testing-strategies

  Use when designing or reviewing tests in the PACT Test phase: test pyramid and coverage, unit/integration/E2E patterns, mocking, counter-test-by-revert, craft pitfalls.

• postgres-supabase

  Use when writing or reviewing Postgres or Supabase schema, migrations, grants, RLS, RPCs, triggers, views, query plans, typegen or drivers (postgres-js, asyncpg, drizzle, pg-boss, FTS): catalog checks over grep, default-privilege and anon traps, proving a policy or constraint fires.

• research-discipline

  Use when research or a source document drives the work: technology, library or API evaluation and comparison matrices, requirements gathering, meeting transcripts and ASR, counting or quoting from a corpus, xlsx/pptx/docx/pdf forensics, verifying a claimed capability, and writing an inventory, audit

• safe-ops

  Use before touching secrets or doing anything destructive: .env and key files (env-get, never source), git init and the first add, Docker and Railway ignore files, Secret Manager values, deleting files, rows, Drive items or branches, and scratchpad cleanup.

• security-review-patterns

  Use when implementing or reviewing auth, credentials or sensitive data, or reviewing code or a design for security: the non-negotiable credential rules (backend proxy only; no secret in frontend code, logs or prompts), sessions, SSRF and credential-bearing fetches, redaction, multi-tenant RLS, launc

• shell-tooling-traps

  Use when shell output is evidence or looks wrong: rtk rewrites if installed (git log/diff, grep, find, tests, lint, helm), test counts via tcount, macOS vs GNU tools, pipes and exit codes, heredoc quoting, long background jobs, sleep, Gatekeeper, and heavy compute (a default cap of 2 workers, nice 1

• stack-traps

  Use when Node/TypeScript or Python backend tooling misbehaves: module resolution, typegen and strictness gaps, monorepo scripts, vitest mocks, pydantic-settings, async event loops, office-document libraries, text normalisation, venvs.

• teammate

  Teammate protocol for repacted specialists. Invoke first at spawn. Read the brief, do the work, hand in with `repact submit`, fix what it names.

• telegram-guide

  Use when telegram_notify is in your tools: when to notify, when to ask and wait, when to check replies, and how several sessions share one chat.

• vendor-gotchas

  Use when integrating third-party APIs not covered by the Postgres or Workspace skills: Monday.com GraphQL, Slack, Jira REST and Forms, Telegram bots, Apify actors, ad, affiliate, outreach and CRM APIs, scraping services, Google client auth off GCP, and deploys on Railway (beside use-railway).

• verification-discipline

  Use BEFORE you report or rely on a finding, number, count, absence ("none", "never", "no callers"), "tests pass", "fixed", "deployed", a relayed claim, or a guard, test or monitor you trust or write. Router of 8 questions with references on artifacts vs descriptions, stale observations, tests that c

• worktree-cleanup

  Use when a PACT worktree is done (PR merged, sub-scope consolidated, manual cleanup): harvest its docs/ artifacts, then remove the worktree and its branch.

• worktree-setup

  Use when starting isolated PACT work: create a git worktree on a feature branch for a new feature, orchestrate or comPACT, or ATOMIZE sub-scope isolation.

Slash commands (15)

• /repacted:autopilot

  Work through the project's task list on its own, one task after another, until the list, the clock or the budget runs out

• /repacted:bootstrap

  Run at session start to reconnect the agent team and show any paused work

• /repacted:comPACT

  Send specialist agents on a small, self-contained task, with no full plan needed

• /repacted:decide

  List every question that waits on you, numbered, and answer a task-log one

• /repacted:demo

  Try rePACTed on a small example project, where one agent fixes one failing test in a scratch copy

• /repacted:init

  Set up this machine and repository for rePACTed, the Telegram bridge, or the CLAUDE.md pins

• /repacted:inspect

  Show where the work stands, or what one task changed and why it was accepted

• /repacted:orchestrate

  Build a feature with a team of AI agents

• /repacted:pause

  Pause the session, or checkpoint it mid-task before /compact, saving memory and state and stopping the helper agents

• /repacted:peer-review

  Commit, open a pull request, and have several agents review it

• /repacted:plan-mode

  Ask several agents for a plan before writing any code

• /repacted:rePACT

  Run a nested plan-design-code-test cycle for a hard sub-task

• /repacted:resume

  Show where the session stands, to pick its work back up after a break

• /repacted:work

  Start work on an ask, on the next task-log row, or on the whole task log unattended

• /repacted:wrap-up

  Clean up at the end of a session and bring the docs up to date

Subagents (12)

• pact-architect

  Designs architecture, component boundaries, API contracts and data flows.

• pact-auditor

  Checks coders' work against the spec as they code; flags drift early.

• pact-backend-coder

  Implements backend code, APIs, business logic and data processing.

• pact-database-engineer

  Implements schemas, migrations, queries, indexes and data integrity.

• pact-devops-engineer

  Implements CI/CD, Dockerfiles, scripts, Makefiles and infrastructure as code.

• pact-frontend-coder

  Implements accessible, responsive user interfaces and their state.

• pact-orchestrator

  Leads the agent team: plans the work, hands it out, checks each result.

• pact-preparer

  Researches docs, APIs and best practices before design or code.

• pact-qa-engineer

  Runs the app to verify pages, interactions and regressions. Needs a dev server.

• pact-secretary

  Saves agents' decisions and lessons; answers project history questions.

• pact-security-engineer

  Reviews code like an attacker and reports vulnerabilities to fix.

• pact-test-engineer

  Designs and runs unit, integration, E2E, performance and security tests.

Hooks (30)

• UserPromptSubmit

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" UserPromptSubmit

• SessionStart

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" SessionStart

• PreCompact

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PreCompact

• PostCompact

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PostCompact

• PreToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PreToolUse Agent

• PreToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PreToolUse Task

• PreToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PreToolUse Read

• PreToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PreToolUse Bash

• PreToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PreToolUse NotebookRead,Grep

• PreToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PreToolUse Monitor

• PreToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PreToolUse WebFetch,Artifact

• PreToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PreToolUse Edit,Write,MultiEdit,NotebookEdit

• PreToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PreToolUse TaskUpdate,TaskCreate

• PreToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PreToolUse AskUserQuestion

• PreToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PreToolUse ScheduleWakeup,CronCreate

• PreToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PreToolUse

• SubagentStop

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" SubagentStop

• Stop

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" Stop

• SubagentStart

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" SubagentStart

• SessionEnd

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/session_end.py"

• SessionEnd

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" SessionEnd

• TeammateIdle

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" TeammateIdle

• PermissionRequest

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PermissionRequest

• PostToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PostToolUse Edit,Write

• PostToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PostToolUse Bash

• PostToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PostToolUse AskUserQuestion

• PostToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PostToolUse TaskCreate

• PostToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PostToolUse TaskUpdate

• PostToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PostToolUse Agent

• PostToolUse

  python3 "${CLAUDE_PLUGIN_ROOT}/hooks/dispatch.py" PostToolUse Read,MultiEdit,NotebookEdit

Similar plugins