claude

codebase-intelligence

Standalone PRP workflow — prp-plan / prp-implement / prp-loop — with Serena LSP structural codebase search, cross-session memory, Jira context injection, a personal KB (local FTS5 vault index), a bounded closed-loop runner with contract-mandated stop rules and an independent verifier, and prp-orchestrate: spec-driven agent teams that groom a Definition-of-Ready spec, audit the whole artifact chain before any code exists, freeze cross-lane contracts, ship one demoable checkpoint per user story, and gate the integrated branch (CI-parity + Copilot/bugbot parity + constitution) before any PR opens. The Obsidian vault is the system of record: run state, pre-pr-gate receipts, the cleanup ledger, plans, reports, session memory and every PR description are vault notes rather than repo-local files, and a failed vault write stops the phase instead of falling back to a copy no sibling worktree or later session can read. Every PR gets its title and body from pr-description automatically — `[TICKET] task description`, written to the vault before the PR is opened. Cross-cutting rules have one copy in shared/ and are cited, not restated, with a validator check that fails on an unresolved citation or a block re-inlined into three or more files. No prp-core required.

1 on GitHubby Artur Gomesv3.30.0MITupdated 3 days ago
/

Plugin codebase-intelligence from marketplace arturgomes in arturgomes/claude-code-toolkit. 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 (35)

  • add-pdf-to-kb

    Ingest PDF or EPUB ebooks into your knowledge base using the bookrag pipeline. Indexes content into a domain-specific bookrag.db (queryable via ask-kb) and creates an Obsidian vault reference note. Use when the user wants to add a book to their KB, process a PDF/EPUB, or expand their knowledge base

  • ask-kb

    Query the personal knowledge base (books + validated principles) for technical or strategic questions. Trigger on "how should I...", "what's the pattern for...", "what does [book] say about...", "consult my KB".

  • benchmark-kb

    Benchmark bookrag retrieval on the obsidian-vault KB; reports MRR/Recall@5/Recall@10/Precision@5 per mode with optional regression check against a saved baseline. Domain: software-craft | functional-programming | all.

  • claude-md-init

    Scaffold a 12-rule CLAUDE.md behavioral contract at the repo root, with attributed rules (Karpathy 1–4 + Mnilax 5–12), an anti-rationalization table, and a tool-routing tail. Trigger on "/init-claude-md", "scaffold a CLAUDE.md", "behavioral contract", "init claude.md".

  • codebase-search

    Serena LSP structural codebase search with session-memory cache-aside. Auto-invoked by prp-plan/prp-implement; trigger manually on "find", "locate", "where is X", "how does Y work".

  • constitution

    Creates and enforces the project constitution — the versioned, non-negotiable architectural principles that outrank any single plan, diff, or deadline. Supplies the three Phase -1 gates (Simplicity / Anti-Abstraction / Integration-First) and the Complexity Tracking table that forces every violation

  • consult-kb

    Review code, RFCs, ADRs, or designs against the personal KB for violations, tensions, and aligned patterns. Trigger on "review this against my KB", "does this follow our patterns?", "audit this design".

  • context7-research

    Fetch version-specific library docs via Context7 MCP before writing any external API call. Auto-invoked by prp-plan/prp-implement; trigger manually on "what's the API for X", "look up docs for Y".

  • doubt-driven

    Mid-flight adversarial review: spawn a fresh-context sub-agent to falsify the implementing agent's strongest claims via grep, then surface mismatches before more code is written. Trigger on "/doubt-this", "challenge my plan", "stress-test this", or auto-invoked from prp-implement once after the last

  • drift-guard

    Anchors every planning and implementation decision to the original AC via seven drift questions. Auto-invoked at every phase gate; invoke manually on "am I drifting?", "is this in scope?", or when a task exceeds expected complexity.

  • function-quality

    Run the 20-item Function Quality Checklist (readability, complexity, DRY, naming, type safety) on specific functions. Trigger after writing a function or on "review this function", "run QCHECKF".

  • index-kb-domains

    Index and wikilink all KB domains in the Obsidian vault so the knowledge base is fully traceable. Creates Obsidian index notes connecting domains → books → core principles. Use when you want to rebuild or refresh the KB navigation layer after adding new books, or to ensure all KB books are reachable

  • ingest-web-doc-to-kb

    Scrape a documentation site or article (a URL and its sub-pages), distill each page into structured KB cards, and index the whole thing into the local bookrag knowledge base — fully autonomously, no API key and no user questions. Mirrors the add-pdf-to-kb pipeline but for the web: discover -> scrape

  • kb-indexer

    Ingest ebooks, PDFs, and documents into the KB as structured markdown cards for use by ask-kb/consult-kb. Trigger on "add this book to my KB", "extract principles from this PDF", "index this ebook".

  • loop-contract

    Defines and validates a Loop Contract (objective, boundaries, executable gate, stop rule, evidence standard, human gate) before any autonomous loop runs. Refuses to loop without an executable binary gate. Auto-invoked by prp-loop Pre-Phase II; invoke manually on "define a loop", "loop contract", "ca

  • mediator

    Coordinator + adversarial judge + merge-gate for /prp-orchestrate. Consumes the plan, slices work into a blocking Foundational round plus one round per prioritized user story, freezes the cross-lane contracts before any worktree forks, assigns disjoint file territory derived from the tasks themselve

  • post-merge-cleanup

    The finish checklist for a shipped unit of work, run AFTER its PR is merged: remove the worktree, delete the branch locally and on the remote, and close the session-memory note with a conclusion. Safety-gated — GitHub's merge state is the authority, nothing local-only is ever destroyed, and a stack

  • pr-description

    The default PR description generator for every pull request this toolkit opens. Derives the title as `[TICKET] task description` (repo root folder name when there is no ticket), writes the body from the branch's real diff — never a placeholder — carries the pre-pr-gate receipt and review fan-out ver

  • pre-pr-gate

    Mandatory mechanical + bot-parity gate that runs on the INTEGRATED branch before any PR is opened. Reproduces locally what CI, GitHub Copilot review, and Cursor bugbot will do to the PR: CI-parity install/typecheck/build/test, a dangling-and-unused-import sweep that CI does not catch, a hygiene swee

  • product-spec

    Generate a structured PRD (user stories, AC, technical constraints) from a feature idea. Trigger on "create a PRD", "define product spec", "act as product owner".

  • prp-constraints

    Feedback loop. Reads the repo's auto-generated CONSTRAINTS.md at the start of every prp-plan / prp-implement run, and appends a new rule whenever a verifier rejects (pr-reviewer, prp-pr-review, pre-pr-gate fail). Rules the system has already learned the hard way never need to be re-learned.

  • prp-pr-review

    Fetch and triage GitHub PR comments via the SKEPTIC framework; applies only valid suggestions with one-line rationale per decision. Run as "/prp-pr-review <PR_URL>" when Copilot or reviewers leave comments on an open PR.

  • quality-review

    Comprehensive quality review: Function Quality (20 items) + Test Quality (16 items) + best practices; produces file:line findings. Trigger after major changes or on "review my code", "run QCHECK".

  • refinement

    Pre-planning grooming gate for /prp-orchestrate. Convenes a scrum-style refinement panel (product-owner + project-manager + lead-engineer + a QA lens) that grooms a goal/ticket/PRD into an unambiguous, no-assumptions contract — prioritized user stories that are each independently testable, numbered

  • session-memory

    Persist and restore findings, decisions, and QA failures to Obsidian vault with BM25 search. Auto-invoked by prp-plan/prp-implement; invoke manually on "save progress", "load context for PROJ-NNN", "resume after QA failure". Provides the Loop Ledger append protocol for prp-loop, and the SESSION CLOS

  • ship

    Scan changes, commit, push, run parallel review fan-out (function/test/security), create a PR, then watch its pipelines and bot reviews until all green — with confirmation at each step

  • skillify

    Extract a reusable SKILL.md draft from a completed plan + report pair in the Obsidian vault. Trigger on "skillify this", "extract a skill from this report", "/skillify <report-path>", or after `prp-implement` finishes when the user wants to cache the pattern for reuse.

  • spec-analyze

    Read-only cross-artifact consistency gate that runs BEFORE any specialist writes code. Grades the refinement contract (spec.md) against plan.md, tasks.md, the territory map, and the constitution — detecting coverage gaps (a requirement with zero tasks), unmapped tasks (scope creep), duplication, amb

  • spec-converge

    Append-only reconciliation of the merged code against the spec, run after the integration gate and before shutdown. Re-reads every FR/SC/acceptance scenario against what the branch actually implements and classifies each gap as missing | partial | contradicts | unrequested, appending the remaining w

  • technical-plan

    Validate implementation plans against codebase patterns for code reuse, DRY, and minimal changes. Trigger on "review my plan", "validate this approach", "does this follow our patterns?".

  • test-quality

    Run the 16-item Test Quality Checklist (structure, coverage, property testing, integration) on test files. Trigger after writing tests or on "review my tests", "run QCHECKT".

  • test-scenarios

    Generate prioritized test scenarios (happy paths, edge cases, errors, performance, security) from a feature description, PRD, or AC. Trigger on "what should I test?", "generate QA scenarios", "test cases for this feature".

  • token-audit

    Audit a Claude Code setup for the 9 token-economy anti-patterns (`claude-code-optimization.md`), produce a structured report with score per pattern and total estimated overhead. Trigger on "/token-audit", "audit token usage", "where are my tokens going", "token economy check".

  • web-search-hook

    Check local web cache before any WebSearch call to avoid redundant requests and token cost. Trigger proactively on "search for...", "find information about...", "what does the web say about...".

  • worktree-lifecycle

    Create, use, and clean up a git worktree for an implementation run. ENTER makes a fresh worktree on a new feature branch off the auto-detected base branch (pulled up to date); EXIT — on user satisfaction — optionally ships (commit/push/PR), saves the session BEFORE deleting, verifies the branch is c

Slash commands (11)

  • /codebase-intelligence:doctor

    Check that all codebase-intelligence prerequisites (system tools, MCP servers, KB engine, vendored tools) are present, and print the exact fix for anything missing.

  • /codebase-intelligence:prp-checkup

    Sweep the PRs you authored that have merged since the last checkup, and finish each one: remove its worktree, delete its branch locally and on the remote, and close its session-memory note with a conclusion. Keeps a ledger so each PR is cleaned exactly once. Merged PRs only — closed-unmerged ones ar

  • /codebase-intelligence:prp-epic

    Run a whole epic to PRs, unattended. Takes a Jira epic key, or an umbrella plan that lists several small plans (and/or tickets) toward one goal. Each child runs as its own detached headless `claude -p` process — a ticket through /prp-orchestrate, a plan through /prp-implement — N at a time in a slid

  • /codebase-intelligence:prp-fix

    Triage a defect as BUG, QA_FAIL or UAT_FAIL and fix it on the short path: no spec, plan or team. New or changed behaviour goes to /prp-plan instead. Pass a JIRA-TICKET, a failure report, or a symptom.

  • /codebase-intelligence:prp-implement

    Implements a .plan.md end-to-end: restores session memory, audits the plan's own requirement→task→gate coverage before the first task, verifies library APIs via Context7 before each task, consults KB for pattern decisions, runs drift-guard before every task, gates the branch with pre-pr-gate, and re

  • /codebase-intelligence:prp-lean

    Lean planning front door for new or changed behaviour: the smallest complete plan, decisions made, at most 3 questions, then a binary GREEN / DOUBT gate. GREEN runs /prp-implement; DOUBT stops with the questions. Defects go to /prp-fix. Pass a goal, a JIRA-TICKET, or path/to/prd.md.

  • /codebase-intelligence:prp-loop

    Runs a bounded, verifiable closed loop: ATTEMPT → GATE → VERIFY → RECORD → DECIDE until an executable objective gate passes (confirmed by an independent verifier sub-agent) or a hard stop fires. Requires a Loop Contract (loop-contract skill) — refuses to run without one. Pass path/to/plan.md (reuses

  • /codebase-intelligence:prp-orchestrate

    Ticket/goal/PRD → done, autonomously. Grooms the input into a spec (prioritized, independently testable stories + FR/SC + a bounded clarify loop), plans it with the full prp-plan rigor (session-memory + Jira injection + codebase agents + ask-kb + Context7 before web + drift-guard + the constitution'

  • /codebase-intelligence:prp-plan

    Transforms a feature description, Jira ticket, or PRD into an implementation plan with session memory, Jira injection, KB consultation, Context7 verification, and drift-guard at every phase gate. Pass a feature description, JIRA-TICKET, or path/to/prd.md.

  • /codebase-intelligence:setup-kb

    Bootstrap the local bookrag KB engine (pinned public base + your own patches) so the KB skills run without any ~/Documents dependency.

  • /codebase-intelligence:sync-vault

    Sync the Obsidian vault repo across machines — commit local changes as a host-stamped snapshot, pull with rebase from origin/main, then push. Run it when opening Claude Code on a machine and again before shutting down.

Subagents (13)

  • backend-specialist

    Generic backend generator for /prp-orchestrate teams. Builds APIs/services/handlers in its OWN git worktree, inside its assigned file territory only. Repo/stack binding (web framework, validation lib) comes from the active preset — NO org specifics here. Consumes core-db contracts; messages qa-analy

  • codebase-analyst

    Traces HOW code works — entry points, data flow, contracts — with Serena LSP for exact file:line resolution and session-memory pre-fill. Use proactively to understand HOW code works, with LSP-verified references.

  • codebase-explorer

    Locates WHERE code lives and extracts implementation patterns via Serena LSP + session-memory pre-fill. Use to find files, map structure, and extract code patterns with file:line references.

  • codebase-researcher

    Autonomous pre-planning research agent; runs session-memory → Serena → structured file:line report. Use when asked to "investigate", "research the codebase for", or "explore before planning".

  • core-db-specialist

    Generic core/database generator for /prp-orchestrate teams. Owns shared types, DB access, and migrations in its OWN git worktree, inside its assigned territory only. Repo/stack + transaction and identifier rules come from the active preset — NO org specifics here. Messages backend-specialist and qa-

  • frontend-specialist

    Generic frontend generator for /prp-orchestrate teams. Builds UI/components/pages in its OWN git worktree, inside its assigned file territory only. Repo/stack binding (framework, component lib) is supplied by the active preset — this agent contains NO org specifics. Messages qa-analyst and ux-specia

  • lead-engineer

    Technical-feasibility lens on the /prp-orchestrate refinement panel. Pressure-tests the goal for buildability: underspecified technical decisions, edge cases, integration unknowns, and the technical Definition of Done. Blocks readiness if the assignment can't be built unambiguously. Advises only — w

  • pr-reviewer

    Generic adversarial reviewer for /prp-orchestrate teams. Performs a harsh, fresh-context review of the merged-candidate diff against the target repo's rule sources (.claude/ + CLAUDE.md + .github/ Copilot instructions, applyTo-scoped) + conventions, one line per finding, severity-tagged. Never the a

  • product-owner

    Business-value owner on the /prp-orchestrate refinement panel. Challenges and authors acceptance criteria, user stories, and scenarios so the goal's BUSINESS intent is captured with zero ambiguity. Blocks readiness if any AC is vague, untestable, or business-obscure. Advises only — writes no code. M

  • project-manager

    Generic planner for /prp-orchestrate teams. Cuts the plan into a small blocking Foundational slice plus one demoable slice per prioritized user story, decomposes each into a granular testable on-disk contract with executable gates + requirement traceability, owns the frozen cross-lane contracts, and

  • qa-analyst

    Generic QA evaluator for /prp-orchestrate teams. Writes test scenarios and runs the contract's behavioral gates against specialist diffs, emitting a pass/fail report. Fresh-context evaluator — never authors the code it grades. Repo/stack binding from the active preset. Messages pr-reviewer.

  • ux-specialist

    Generic design-taste evaluator for /prp-orchestrate teams. Runs a before/after taste check on UI-affecting diffs against a design/originality/craft/functionality rubric. Advises frontend; does not author feature code. Repo/stack binding from the active preset. Messages frontend-specialist.

  • web-researcher

    Web researcher with KB + Context7 pre-checks (covered topics skip web) and drift-guard question #5 on findings. Use when information is beyond training data — KB and Context7 run first to minimize unnecessary web calls.

Hooks (8)

  • SessionStart

    bash "${CLAUDE_PLUGIN_ROOT}/scripts/on-session-start.sh"

  • PreToolUse

    bash "${CLAUDE_PLUGIN_ROOT}/scripts/run-node-hook.sh" "${CLAUDE_PLUGIN_ROOT}/hooks/scratch-guard/guard-scratch-md.js"

  • PreToolUse

    bash "${CLAUDE_PLUGIN_ROOT}/scripts/run-node-hook.sh" "${CLAUDE_PLUGIN_ROOT}/hooks/adversarial/mark-adversarial.js"

  • PostToolUse

    bash "${CLAUDE_PLUGIN_ROOT}/scripts/run-node-hook.sh" "${CLAUDE_PLUGIN_ROOT}/hooks/adversarial/track-edit.js"

  • PostToolUse

    bash "${CLAUDE_PLUGIN_ROOT}/scripts/run-node-hook.sh" "${CLAUDE_PLUGIN_ROOT}/hooks/adversarial/track-bash-edit.js"

  • SubagentStop

    bash "${CLAUDE_PLUGIN_ROOT}/scripts/run-node-hook.sh" "${CLAUDE_PLUGIN_ROOT}/hooks/adversarial/subagent-adversarial.js"

  • Stop

    bash "${CLAUDE_PLUGIN_ROOT}/scripts/run-node-hook.sh" "${CLAUDE_PLUGIN_ROOT}/hooks/adversarial/require-adversarial.js"

  • UserPromptSubmit

    bash "${CLAUDE_PLUGIN_ROOT}/scripts/run-node-hook.sh" "${CLAUDE_PLUGIN_ROOT}/hooks/adversarial/reset-turn.js"

View source on GitHub

More from arturgomes/claude-code-toolkit

All 2 plugins →

Turns a lesson, text or topic into an image prompt for an educational infographic for school students (early/middle/high), 18 layouts x 15 styles, text in pt-BR or en.

1 skill

1

Similar plugins

Personalized AWS guidance built on patterns from 350,000+ startups. Get architecture, cost, and security recommendations matched to your stage, from day-one account setup to production-ready infrastructure. Migrate to AWS from Azure, GCP, or Heroku with resource discovery, cost estimates, and generated Terraform. Move OpenAI, Azure OpenAI, and Gemini code to Amazon Bedrock, and plan and prototype AI agents on AgentCore and other AWS runtimes. Includes AWS Activate credits eligibility, 60+ exclusive startup offers, and multi-account, multi-region support. Built by AWS Startup Solutions Architects.

13 skills · 1 command · 7 agents · 1 MCP server

anthropics/claude-plugins-official 47,076

gbrain

claudecodex

Personal knowledge brain for your coding agent — hybrid search, synthesis, and durable cross-session memory, plus a curated brain-first skill set.

80 skills · 1 MCP server · hooks

garrytan/gbrain 32,792

Converts books, documentation folders, and source collections (PDF, EPUB, DOCX, HTML, Markdown, RST, AsciiDoc, RTF, MOBI/AZW) into structured agent skills: a master SKILL.md with core frameworks and a topic index, on-demand chapter files, a glossary, a patterns file, and a decision cheatsheet. Ships 4 stdlib-only tools (multi-format extractor with invisible-Unicode sanitization, four-family generated-skill validator covering frontmatter/safety/budget/index, pre- and post-flight token budget estimator with a worth-converting verdict, and a plugin emitter that wraps a compiled skill as an installable claude-skills package behind a rights gate), 5 references citing 7-8 sources each, a cs-book-to-skill agent, and /cs-book-to-skill + /cs-book-to-plugin commands. Derived from virgiliojr94/book-to-skill (MIT).

1 skill · 2 commands · 1 agent

alirezarezvani/claude-skills 29,725

Answer technical questions using daily.dev's knowledge base. Searches articles from the developer community and synthesizes answers with source links.

6 skills

dailydotdev/daily 20,137

Local-first Agent Skills package and Claude Code plugin for source-cited Obsidian knowledge bases. Includes recoverable transactions, deterministic lint and retrieval, provenance ledgers, explicit vault adoption, offline-first capture foundations, and methodology modes.

15 skills · 3 agents · hooks

AgriciDaniel/claude-obsidian 16,385

Skills for local knowledge-base retrieval and document QA.

1 skill

ConardLi/garden-skills 15,041

Maintain codebase-intelligence?

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 plugins2 plugins