From kadai
Use when verifying that kadai's full integration (skill auto-trigger + PreToolUse/PostToolUse hooks + MCP server + change capture + status updates) works end-to-end. Spawns a fresh `claude -p` session in a temp dir to exercise the Path A acceptance test from the kadai design spec §10. Useful after kadai upgrades, install changes, or when something feels off about hooks/MCP/skill behavior.
How this skill is triggered — by the user, by Claude, or both
Slash command
/kadai:kadai-dogfood-testThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
This skill encapsulates the formal Path A dogfood test for kadai. It spawns a fresh Claude Code session inside a temp dir and observes whether the agent uses kadai correctly, whether hooks fire, and whether change capture works.
This skill encapsulates the formal Path A dogfood test for kadai. It spawns a fresh Claude Code session inside a temp dir and observes whether the agent uses kadai correctly, whether hooks fire, and whether change capture works.
Use it after kadai upgrades, after editing the plugin / skill / hooks / MCP tools, or when debugging why the kadai workflow doesn't seem to be firing as expected.
From kadai design spec §10:
kadai skill fires on planning/implementation language before any code is written.Edit/Write outside the spine.kadai pick STORY-001 (or mcp__kadai__pick_story), status moves to in_progress, next edit succeeds.changelog.md.mcp__kadai__set_status to mark the story review (or done) when work completes.kadai serve (skipped by default in this skill — start it separately if you want to verify).TMP=$(mktemp -d -t kadai-dogfood-XXXXXX)
echo "Test temp dir: $TMP"
cd "$TMP"
kadai init -y > /dev/null
kadai add epic --title "Test Epic" --phase mvp > /dev/null
kadai add feature --title "Math utils" --phase mvp --parent EPIC-001 > /dev/null
kadai add story --title "Implement add(a,b)" --phase mvp --parent FEAT-001 > /dev/null
# Verify spine
kadai list epic && kadai list feature && kadai list story
claude -p session in the temp dircd "$TMP"
echo "STORY-001 needs implementing — 'Implement add(a,b)'. Use the kadai workflow: pick the story, then create src/math.ts with an add(a, b) function and tests/math.test.ts using bun:test. Mark the story 'review' when done. Report which kadai commands you used and any hook errors you saw." \
| claude -p --model haiku \
--allowedTools "Bash Edit Write Read mcp__kadai__set_status mcp__kadai__pick_story mcp__kadai__get_active_story mcp__kadai__list_stories mcp__kadai__get mcp__kadai__create_task" \
2>&1 | tee /tmp/kadai-dogfood-output.txt | tail -40
The --allowedTools whitelist is conservative — only what the agent needs. Add tools as the test evolves.
cd "$TMP"
echo "=== (point 1+3) Did the agent's report mention kadai pick / kadai workflow? ==="
grep -iE "(pick|kadai)" /tmp/kadai-dogfood-output.txt
echo "=== (point 4) Changelog populated? ==="
find .kadai -name changelog.md -exec echo "{}:" \; -exec cat {} \; -exec echo "" \;
echo "=== (point 5) Story status moved? ==="
kadai list story
echo "=== Created files (sanity) ==="
ls src/ tests/ 2>&1
echo "=== Final overall status ==="
kadai status
For each of the 6 points, mark ✅/❌/SKIPPED with the evidence above:
kadai pick STORY-001 was calledreview or donekadai serve separatelyVerdict: PASS / PASS-with-notes / FAIL.
If the test passed (or had notable findings), append a dated entry to docs/dogfood-acceptance-test.md in the kadai repo so we have a history of dogfood runs over time.
rm -rf "$TMP" /tmp/kadai-dogfood-output.txt
--model haiku for speed. Switch to sonnet if you want more deliberation from the test agent.claude isn't on PATH: the test can't run. The skill prerequisite is a working claude CLI.mcp__kadai__set_status is in --allowedTools. The skill's allowlist already includes it; if you've modified the command, re-check.claude install — this skill assumes you can spawn claude -p locally.Provides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub fintanislost/kadai --plugin kadai