core-skills

Plugin core-skills from marketplace script-development-plugins in script-development/claude. Commands read from the repo's manifest — 14 days ago.

Skills (15)

• catchup

  Load the full context of the current branch and check alignment with the base: git diff summary, commit history, plan/task files, decisions, and (if the project has an issue-tracker skill) the upstream issue details. Reports how far behind the base the branch is and offers to merge on confirmation.

• commit

  Create small, focused commits with proper messages and push to remote. Matches this project's own commit-message convention rather than imposing one, and reads .claude/project-context.md for this project's integration branch so the protected-branch check doesn't miss a non-default name. Keeps the br

• fix-bug

  Structured bug-fix workflow: identify the issue, reproduce the defect, confirm it isn't already fixed on the base branch, capture the investigation in docs/bugs/<slug>/BUG.md, propose the fix and wait for developer approval, implement it, gate on the bug-fix-verifier agent, and gate styling/animatio

• implement-plan

  Implement a feature directly from PLAN.md and DECISIONS.md without a TASKS.md breakdown. Use this for smaller plans where the plan's Approach section is already a good enough roadmap and a separate task file would be overkill. Loads the full plan context (goal, decisions, scope, approach, acceptance

• install

  Create or top up this project's `.claude/project-context.md` — the one file every installed plugin skill (core-skills, kendo-pm, ...) reads for project-specific facts — from the shipped template, keeping only the fields the skills actually installed here read. Detects values from the repo first and

• newbranch

  Create a new branch from this project's integration branch. If this project has an issue tracker configured, resolve or create the issue first and prepare it for work (assign, add to the active sprint, move to In Progress, link the branch). Use whenever the user wants to start a new feature, create

• next

  Continue working through TASKS.md — find the next uncompleted task, execute it with TDD flow, and mark it done. Use this skill whenever the user wants to continue implementation work, pick up where they left off, resume after /clear, or work through a task list. Trigger on: "next task", "continue wo

• plan-feature

  Structured feature planning workflow: interrogate the developer with hard, codebase-informed questions until requirements are crystal clear, then produce PLAN.md and DECISIONS.md. Use when planning a feature, designing a feature, "how should we build this", "let's plan", "I want to build X", or any

• pr

  Create a pull request for the current branch targeting the base branch, with automatic issue feedback (when the branch links to a tracked issue and this project has an issue tracker configured). Offers to run `/core-skills:review-branch` when this session has no review for HEAD, gates bug branches o

• review-branch

  Full-branch review of the current branch vs the integration branch. Spawns `runtime-integrity-reviewer`, `correctness-reviewer` and `precedent-reviewer` in parallel and reports in chat for `/core-skills:pr` to consume in this session. Use whenever the user wants to review a branch, says "review my c

• review-mcp-descriptions

  Review and improve MCP tool and resource descriptions for Tool Search discoverability. Use after adding or modifying MCP tools or resources to ensure descriptions follow best practices. Trigger on: "review mcp descriptions", "check tool descriptions", "optimize mcp tools", or after creating new MCP

• shepard

  Drive ONE open pull request to green and answered. Each cycle reads red CI and the reviewer's findings, gives every row a disposition, fixes what is unambiguous, grills the developer on the design calls, replies in every thread, pushes once, and arms a live watch on the PR. Use it whenever the user

• sync-worktrees

  Sync every secondary git worktree with the primary one. Copies every gitignored `.env*` file from the primary into each secondary worktree at the same relative path, and refreshes dependencies (composer install where a composer.json exists, npm install where a package.json exists). Optionally fast-f

• task-writer

  TDD task breakdown and TASKS.md management. Use when planning a new feature, breaking down work into implementation tasks, creating or updating TASKS.md, estimating effort, or organizing multi-step work into phases. Also use when translating a plan document (like docs/plans/*.md) into TASKS.md — inc

• worktree

  Cut a fresh git worktree in the current repo and set it up — branch, dependencies, env files, ready to work. Project-agnostic: auto-detects the integration branch and the setup steps, and reads .claude/project-context.md's Worktrees section for this project's own overrides and house rules when one e

Subagents (6)

• bug-fix-verifier

  Verify that a bug fix actually resolves the defect described in BUG.md and doesn't introduce obvious regressions. Spawned by `/core-skills:fix-bug` before PR creation as the blocking gate for bug-fix branches.

• correctness-reviewer

  Review a branch for code that computes the wrong thing, paths the happy-path test never takes, and obligations the change created but did not meet — including user-facing text that promises behaviour the code does not deliver. Carries the Runtime-state simulation, Sins of omission, and Test discipli

• plan-reviewer

  Review feature plans for codebase convention violations before approval. Use when a plan has been produced by /core-skills:plan-feature and needs convention checking, or when the task mentions plan review, convention check, or plan audit.

• precedent-reviewer

  Review a branch against what is already written down — the repo's standing rules and architecture decisions, the sibling implementation that already solves this shape, the branch's own plan prose, and the CI configuration that decides what a green check means. Carries the Recorded rulings, Sibling p

• runtime-integrity-reviewer

  Review a branch for failures that vanish and guards that got weaker — swallowed errors, partial completion, boundaries crossed without a check, entry points missing a guard, and security posture the diff loosened. Carries the Security & cost surface, Runtime data-flow, and Contract & boundary corpus

• surface-reviewer

  Review a plan's Security & Cost Surface against seven question-shaped rows: untrusted input → LLM, external mutation partial-failure space, endpoint surface (authz / rate / state / cross-tenant), audit-log fidelity, state-machine walkthroughs with degradation signals, convention enforcement level, a

Similar plugins