From autoforge
Capability map for AutoForge as a durable runtime spanning software development, academic research, verification, daemonized execution, async interruption, and eval harnesses. Use this to decide when another AutoForge plugin skill should be applied.
How this skill is triggered — by the user, by Claude, or both
Slash command
/autoforge:autoforge-runtime-overviewThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
AutoForge is most useful when a task benefits from a durable multi-agent runtime rather than a single short Claude Code interaction.
AutoForge is most useful when a task benefits from a durable multi-agent runtime rather than a single short Claude Code interaction.
Prefer AutoForge when the user needs one or more of:
Core runtime primitives:
autoforgeai daemon startautoforgeai queue ...autoforgeai watch <project_id>autoforgeai msg <project_id> "..."autoforgeai unpause <project_id>autoforgeai harness run ...autoforgeai harness openai-export ...autoforgeai paper infer|reproduce ...Decision rules:
long-run-runtime skill.software-forge skill.academic-repro skill.harness-evals-export skill.npx claudepluginhub alyciabhz/autoforge --plugin autoforgeExecutes long-running tasks autonomously across Claude Code sessions using headless bash loops or in-session hooks. Supports structured decomposition for projects and Ralph-style iteration for TDD, fixes, refactoring.
Orchestrates complex tasks with multi-agent Planner → Generator → Evaluator → Retro pipeline, clean-context sub-agents, drift prevention, and persistent retro learning. Use for 'harness this task' or structured coordination.
Designs autonomous agent harnesses with research loops, evaluation scaffolds, locked/editable surfaces, durable logs, novelty gates, pruning, rollback, and human approval boundaries.