paradigm

Plugin paradigm from marketplace a-paradigm in ascend42/a-paradigm. Commands read from the repo's manifest — today.

Skills (21)

• agents

  Manage your agent team — roster, onboard, bench/activate, detail, install from GitHub or nevr.land. Use when the user says "agents", "show my agents", "onboard agents", "install agent", "bench agent", or wants to manage their AI team.

• class

  Convene the Classroom — the gated, teacher-led learning term. Agents bring proposed learnings; peers refute them with test-case scenarios; you sign off, and survivors promote to notebooks (or get refined). Use when the user says "class", "hold class", "convene the classroom", "review what the agents

• conduct

  Register this Claude Code session with Paradigm Conductor. Use when the user says "conduct", "surface to conductor", "register with conductor", or wants this terminal visible in the Conductor overlay.

• doctor

  Run a comprehensive health check on the Paradigm setup. Use when something seems wrong, after major changes, or when the user says "check paradigm health", "paradigm doctor", or "is paradigm set up correctly".

• forage

  Send an agent on an EXPEDITION into the wild — Reddit, Medium, engineering blogs, Anthropic docs — to grow its knowledge around a topic. Returns CITED candidate learnings (never certifies), staged at the floor trust tier for the gated class to adjudicate. Nothing schedules it by default: it runs whe

• handoff

  Prepare a context handoff summary for the next session or agent. Use when the user says "prepare handoff", "I'm done for now", "hand this off", or when context usage is high and work needs to continue later.

• health

  Agent learning health dashboard — Neverland metrics, nomination stats, threshold drift, notebook growth. Use when the user says "health", "agent health", "neverland", "learning metrics", "how are agents doing", or wants to check the learning system.

• init

  Initialize Paradigm in the current project. Use when a user says "set up paradigm", "initialize paradigm", "add paradigm to this project", or when you detect a project has no .paradigm/ directory.

• lore

  Record a lore entry for the current session. Use when the user says "record lore", "log what we did", "save session history", or proactively at the end of a significant work session (3+ files modified).

• observe

  View live logs, metrics, and traces from Sentinel. Use when the user says "show logs", "check metrics", "view traces", "what's happening", or when monitoring application behavior.

• postflight

  Run post-task compliance checks after finishing implementation. Use when done with a task, when the user says "postflight", "am I done", "check my work", or proactively before ending a session where source files were modified.

• preflight

  Run pre-task compliance checks before starting implementation. Use when starting a new task, when the user says "preflight", "before I start", "what should I check", or proactively before complex tasks that affect 3+ files.

• protocol

  Search for or record a repeatable implementation protocol. Use when the user says "check protocol", "record protocol", "any protocol for this", or before starting implementation of a common pattern.

• review

  Review recent changes for Paradigm compliance and code quality. Use when the user says "review my changes", "check my code", "am I good to commit", or after implementation before committing.

• ripple

  Analyze the impact of changing a symbol before modifying it. Use when the user says "what would break", "impact analysis", "ripple check", or before making changes to a core symbol.

• scan

  Rebuild the Paradigm symbol index after file changes. Use when symbols seem out of date, after creating or moving files, or when search results are stale.

• sentinel

  Triage and manage incidents using Paradigm Sentinel. Use when the user says "check incidents", "triage errors", "sentinel status", "what errors happened", or when investigating production issues or test failures.

• shift

  Full one-command Paradigm setup — initializes everything, scans symbols, installs hooks, generates CLAUDE.md, and runs doctor. Use when a user wants complete setup in one step, says "paradigm shift", "set up everything", or "full paradigm setup".

• study-hall

  Autonomous self-study — each active agent follows its approved curriculum, drills against the scenario bank, and STAGES candidate learnings for the next gated class. Never certifies, never writes notebooks; drills never count as evidence. Nothing schedules it by default: it runs when the human start

• teach

  Teach an agent a new behavior or pattern. Writes a journal entry that promotes to the agent's notebook for future sessions. Use when the user says "teach", "train agent", "agent should know", "remember this for agent", or wants to give an agent specific instructions.

• team

  Show what your agent team did this session — contributions, verdicts, learning results. Use when the user says "team", "what did the team do", "session summary", "who contributed", or wants to see agent activity.

Subagents (5)

• architect

  System design and architecture planning agent. Use when tasks require architectural decisions, designing multi-file features, choosing between implementation approaches, or planning changes that span 3+ components. Read-only — does not write implementation code.

• builder

  Implementation and coding agent. Use for writing code, creating files, and making changes based on specifications or task descriptions. Follows existing patterns, updates .purpose files alongside code, and uses the Paradigm logger instead of console.log.

• reviewer

  Code review and quality analysis agent. Use to review changes for bugs, spec compliance, code smells, and convention adherence. Read-only — does not modify code. Uses a two-stage protocol: spec compliance first, then code quality.

• security

  Security analysis agent for authorization, authentication, input validation, and vulnerability detection. Use when tasks involve auth, user data, API endpoints, or when reviewing code for security issues. Read-only — flags issues for Builder to fix.

• tester

  Testing and validation agent. Use after implementation to write tests, run test suites, verify edge cases, check gate validations, and ensure code quality. Can read and write test files, and run test commands.

MCP servers (1)

• paradigm

  runs: node ${CLAUDE_PLUGIN_ROOT}/scripts/paradigm-mcp-launch.cjs ${CLAUDE_PROJECT_DIR}

Hooks (9)

• SessionStart

  ${CLAUDE_PLUGIN_ROOT}/scripts/paradigm-session-start.sh

• Stop

  ${CLAUDE_PLUGIN_ROOT}/scripts/paradigm-stop.sh

• PreToolUse

  ${CLAUDE_PLUGIN_ROOT}/scripts/paradigm-precommit.sh

• PreToolUse

  ${CLAUDE_PLUGIN_ROOT}/scripts/paradigm-navigate-remind.sh

• PreToolUse

  ${CLAUDE_PLUGIN_ROOT}/scripts/paradigm-team-gate.sh

• UserPromptSubmit

  ${CLAUDE_PLUGIN_ROOT}/scripts/paradigm-prompt-gate.sh

• PostToolUse

  ${CLAUDE_PLUGIN_ROOT}/scripts/paradigm-postwrite.sh

• PostToolUse

  ${CLAUDE_PLUGIN_ROOT}/scripts/paradigm-skill-used.sh

• PostToolUse

  ${CLAUDE_PLUGIN_ROOT}/scripts/paradigm-delegation.sh

Similar plugins