How this skill is triggered — by the user, by Claude, or both
Slash command
/vader:execThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Execute the current vader plan through a ralph-wiggum loop.
Execute the current vader plan through a ralph-wiggum loop.
Run the permissions check:
"${CLAUDE_PLUGIN_ROOT}/scripts/check-permissions.sh"
If the output is NOT bypassPermissions, warn the user:
Vader works best with
--dangerously-skip-permissionsto avoid permission prompts interrupting the loop. Consider restarting with:claude --dangerously-skip-permissions
Ask if they want to continue anyway. If not, stop.
Read the plan file:
.claude/vader/plan.local.md
If the file does not exist, tell the user to run /vader first and stop.
Run the setup script to compose the ralph-wiggum prompt:
"${CLAUDE_PLUGIN_ROOT}/scripts/setup-exec.sh"
The output is the composed prompt text directly.
Read the max_iterations from the plan file frontmatter.
Try to invoke the ralph-wiggum loop using the Skill tool:
ralph-loop:ralph-loop<prompt-from-step-3> --max-iterations <max_iterations> --completion-promise 'Hurra Vader has Triumphed'Drop to direct execution on any failure of the ralph-loop invocation: skill not installed, prompt parse error, runtime error, anything. When falling back, tell the user explicitly:
Falling back to direct execution (ralph-loop unavailable: ). The same per-milestone Executor → Verifier flow runs inline; only the iteration controller differs.
Don't silently retry. The failure mode of ralph-loop is opaque to the user; surfacing the reason makes it possible to fix.
In direct execution mode:
.claude/vader/plan.local.md for the full plantest command is)npm run test:sanity, etc.)is_final: true<promise>Hurra Vader has Triumphed</promise>The direct execution mode works identically to the ralph-wiggum loop but runs inline in the current session.
Guides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.
npx claudepluginhub kobbikobb/vader --plugin vader