foundry-core

Plugin foundry-core from marketplace foundry in fedcal/foundry. Commands read from the repo's manifest — 1 month ago.

Skills (8)

• audit-agent-surface

  Audit the coding agent's own configuration as an attack surface — permission rules and mode, every hook command and the writability of the script it executes, MCP servers and what they expose, installed plugins and marketplace provenance, and credentials reachable from any of them. Treats settings.j

• contracts

  Work with Foundry I/O contracts — validate an artifact, choose the right schema for a handoff, or add a new contract version. Use when writing to the blackboard, when a contract validation error appears, or when designing a new agent's output.

• foundry-init

  Bootstrap Foundry in a project — create the .foundry state directory, wire settings and CLAUDE.md, pick a profile, and seed memory from what already exists. Use when a project has no .foundry directory yet, when `foundry doctor` reports missing state, or when the user asks to set up Foundry here.

• handoff

  Hand work from one agent or session to the next without dumping context — write the artifact, summarise inside budget, and state what is unresolved. Use when finishing a delegated task, ending a wave, or pausing work that someone else will resume.

• memory

  Capture, recall, curate and prune Foundry project memory. Use when a decision or constraint has just been established, when you need to know what was decided before, or when the memory index exceeds its token budget.

• orchestrate

  Run multi-agent work in waves against a Foundry playbook — delivery, audit, research or migration. Use when a task needs several specialists working in parallel with gates between stages, or when the user asks to orchestrate, fan out, or run a full cycle.

• runbook

  Find, follow, create or revise an operational runbook. Use BEFORE starting any recurring or error-prone task (deploy, release, migration, incident, audit), and immediately AFTER finishing one that someone will repeat.

• token-budget

  Measure and reduce what a Claude Code session costs — where tokens go, which levers actually move the number, and how much Foundry's memory system is saving. Use when a session feels expensive, before starting long work, or when the user asks about cost, token usage or context pressure.

Subagents (4)

• context-broker

  Retrieves the minimum context needed to answer a question or start a task — searches memory, runbooks, ADRs and the codebase, then returns a compact briefing instead of file dumps. Use before planning or implementing when you would otherwise read many files to orient yourself.

• foundry-orchestrator

  Plans and runs multi-agent work in waves against a Foundry playbook. Use when a task needs more than one specialist, when work can be parallelised across independent areas, or when the user asks for a full delivery, audit or research cycle. Coordinates specialists, enforces contracts between waves a

• memory-curator

  Curates Foundry project memory — turns a caller-supplied list of candidate facts into stored `fact.v1` entries, deduplicates them, retires what is no longer true and keeps the index inside its token budget. Use at the end of a working session, before compaction, or when the memory index has grown pa

• runbook-author

  Writes and maintains operational runbooks after work that will recur — deploys, incidents, migrations, releases, recurring audits. Use immediately after finishing a task that was error-prone or that someone will have to repeat, and whenever an existing runbook proved wrong or incomplete. Put the com

MCP servers (1)

• foundry

  runs: node ${CLAUDE_PLUGIN_ROOT}/mcp/server.mjs

Hooks (9)

• SessionStart

  node

• UserPromptSubmit

  node

• PreToolUse

  node

• PreToolUse

  node

• PostToolUse

  node

• SubagentStop

  node

• Stop

  node

• PreCompact

  node

• SessionEnd

  node

Similar plugins