src/.agents/skills/bootstrap/SKILL.md
Prepare the assigned spec worktree and local validation environment before any execution role begins.
npx skillsauth add tolulawson/ralph-harness bootstrapInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
implement or another worker role is about to run in a session that has not yet passed bootstrap for that claim..ralph/shared/state/spec-queue.json or the resolved canonical .ralph/state/spec-queue.json.ralph/shared/state/execution-claims.json or the resolved canonical .ralph/state/execution-claims.json.ralph/shared/context/project-truths.md or the resolved canonical .ralph/context/project-truths.md.ralph/shared/context/project-facts.json or the resolved canonical .ralph/context/project-facts.json.ralph/shared/context/learning-summary.md or the resolved canonical .ralph/context/learning-summary.md.ralph/shared/policy/project-policy.md or the resolved canonical .ralph/policy/project-policy.mdgit status --short --branch.ralph/context/project-facts.json, unless the spec queue entry explicitly overrides base_branch..ralph/worktrees/..ralph/shared/ resolves shared artifacts back to the canonical checkout..ralph/context/project-facts.json bootstrap_env_filesbootstrap_copy_exclude_globsnode_modules, virtualenvs, .next, dist, build, .turbo, .cache, and similar paths alone unless the project facts explicitly opt into a narrower policy.ralph/context/project-facts.json worktree_bootstrap_commands.ralph/context/project-facts.json validation_bootstrap_commands.ralph/context/project-facts.json verification_commands when they are safe as bootstrap checks.ralph/state/execution-claims.json so the claim records whether bootstrap is in_progress, passed, or failed..ralph/reports/<run-id>/<spec-key>/bootstrap.md, typically via .ralph/shared/reports/.implement only when bootstrap passed and the claim is validation-ready..ralph/shared/ overlay inside the assigned worktree.ralph/state/execution-claims.json.ralph/reports/<run-id>/<spec-key>/bootstrap.mdStop after the assigned claim has a terminal bootstrap result, the bootstrap report is written, and the next role is clear.
development
Lightweight pre-review quality gate that removes common slop from changed code before handoff. Focuses on type strictness, single-responsibility boundaries, fail-fast error handling, DRY cleanup, and dead-code/workaround removal.
testing
Perform a read-only cross-artifact consistency analysis across a numbered spec's spec, research, plan, and tasks files, surfacing duplication, ambiguity, coverage gaps, and inconsistencies before implementation continues.
testing
Upgrade an already-installed Ralph multi-agent runtime in the current repository from the latest tagged scaffold surface while preserving project-owned runtime files and history.
tools
Generate or update a Ralph-style project PRD in the current repository by asking a small number of high-value clarifying questions, then writing a structured PRD with epochs, bounded user stories, and verifiable acceptance criteria.