agents-core

Plugin agents-core from marketplace agents-template in toderian/project_template. Commands read from the repo's manifest — 2 days ago.

Skills (18)

• codebase-design

  Shared vocabulary and principles for designing deep modules: a lot of behaviour behind a small interface, placed at a clean seam, testable through that interface. Use when designing or reviewing a module's interface, deciding where a seam goes, hunting for deepening opportunities, making code more t

• connectors-and-mcp

  Least-privilege rules for connectors and MCP servers (GitHub, Google Drive, Slack, Linear, databases, internal APIs, cloud consoles) by autonomy level, with secrets handling and stop points. Use before reading/writing a live external system via a connector/MCP server, or when choosing which one.

• diagnose

  Disciplined diagnosis loop for hard bugs and performance regressions: reproduce -> minimise -> hypothesise -> instrument -> fix -> regression-test. Use when the user says \"diagnose this\" / \"debug this\", reports a bug, or describes broken/failing or slow behavior.

• execute-plan

  Execute an approved task or implementation plan phase-by-phase with a committed run state on disk: small tasks are implemented in the main thread with one independent reviewer per phase, large ones as a thin orchestrator dispatching a fresh implementer and separate spec, quality and security reviewe

• git-discipline

  Branch, commit, squash and push rules plus the autonomy ladder (L0-L3). Use when deciding where to commit, whether to branch, how to write commit messages, when squashing task commits, or when a push/force/reset is being considered.

• grilling

  Grill the user relentlessly about a plan, decision, or idea until every branch of the design tree is resolved. Use when the user wants to stress-test their thinking, says \"grill me\", or another skill needs decisions only the user can make.

• handoff

  Compact the current conversation into a handoff document so a fresh agent can pick up the work. Use when the user wants a session summary written to disk for later continuation, mentions \"handing off\", or is wrapping up a long session.

• performance-optimization

  Measure-first performance optimization: profile the real hot path, record a baseline, change one thing, and re-measure. Use when the user wants to optimize performance, reduce latency/memory, speed something up, or asks \"why is this slow?\".

• planning-workflow

  Seven-step pre-implementation planning workflow with a five-axis adversarial critique rubric, including tracer-bullet vertical slicing for plans built from a PRD. Use when the user wants to plan a non-trivial change before writing code, break a PRD into implementation phases, or mentions \"tracer bu

• prototype

  Build a throwaway prototype before committing to a design: a runnable terminal app, or several UI variations. Use when the user wants to prototype, test a data model/state machine, or says \"prototype this\", \"let me play with it\", or \"try a few designs\".

• security-review-owasp

  Apply current OWASP standards (Top 10, ASVS, LLM Top 10, Agentic AI) when writing or reviewing code. Use when reviewing code for security issues, implementing auth/authz, handling user input, designing API endpoints, building AI agent systems, or integrating LLMs/RAG.

• setup-project

  Set up or update a repository's agent contract: seed it with `at init`, fill AGENTS.md's Project section with the human, and adopt template updates after the plugins move. Use when setting up a new repo, after pulling or updating the plugins, when `at doctor` reports routing-table drift, or when AGE

• simplicity-review

  Review code, plans, dependencies, or repos for avoidable complexity: reuse local code, prefer stdlib/native, avoid speculative abstractions. Use when the user asks for a simplicity, over-engineering, or YAGNI review, minimal implementation planning, what can be deleted/replaced with stdlib, or a sim

• spec-workflow

  Spec-driven plan-build-review-fix loop for a non-trivial item, producing specs/<slug>/ artifacts. Use when the user asks to \"spec it out\" / \"spec workflow\" / \"spec-driven\", or needs plan+build+review too large for the default loop; not trivial edits or spikes.

• squash-workspace-commits

  Squash completed workspace/task commits after validation while protecting unrelated or pushed history. Use when the user says \"squash commits\", \"clean up commits\", \"squash task commits\", or asks to compact agent phase commits after review.

• subagent-protocol

  Multi-agent coordination protocol with status vocabulary, dispatch format, and two-stage review. Use when the user wants to dispatch subagents, coordinate multi-agent work, or review implementation output.

• tdd

  Test-driven development with red-green-refactor loop. Use when the user wants to build features or fix bugs using TDD, mentions \"red-green-refactor\", wants integration tests, or asks for test-first development.

• wait-what

  Stop — that last message did not land: re-pitch it.

Subagents (6)

• implementer

  Implements a single task slice from a plan. Use when dispatching focused implementation work to a subagent.

• plan-critic

  Adversarial review of a plan before implementation. Scores against five axes (assumption audit, scope creep, existing solutions, minimalism, uncertainty) and issues PROCEED, REVISE, or BLOCKED. Use after a plan is drafted and before code is written.

• researcher

  Investigate before recommending. Searches codebase first, then authoritative external sources. Returns evidence-backed findings with citations and tradeoffs. Use when the team needs grounding before action — evaluating libraries, comparing approaches, surfacing prior art, refreshing doctrine.

• reviewer

  Two-stage review of implementation work. Use after an implementer completes a task to verify spec compliance and code quality.

• security-auditor

  Security review of implementation changes. Applies OWASP Top 10 + LLM Top 10 + Agentic AI checks via the agents-core:security-review-owasp skill. Distinguishes real vulnerabilities from correctly-managed configuration. Use after implementation, before merge.

• spec-validator

  Spec-blind behavioral validation. Reads only the acceptance criteria, writes binary pass/fail tests against the implementation, and reports a shared subagent status plus PASS/FAIL verdict. Provides an independent verification layer that cannot be biased by seeing how the code was built.

Hooks (7)

• PreToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/block-dangerous-git.sh

• PreToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/block-dangerous-bash.sh

• PreToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/block-write-sensitive.sh

• PreToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/block-bad-todo-name.sh

• PostToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/remind-archive-done-todo.sh

• SessionStart

  "${CLAUDE_PLUGIN_ROOT}"/hooks/session-drift.sh

• SubagentStop

  "${CLAUDE_PLUGIN_ROOT}"/hooks/require-subagent-status.sh

Similar plugins