kiln

Plugin kiln from marketplace yoshisada-kiln in yoshisada/ai-repo-template. Commands read from the repo's manifest — 1 month ago.

Skills (43)

• task-shapes
• audit

  Run a PRD compliance audit against the current implementation. Called automatically at the end of /implement. Checks both directions: PRD→Spec and Spec→Code→Test. Attempts to fix gaps, or requires documented blockers.

• implement

  Execute all tasks from the task breakdown to build the feature. Use after

• kiln-analyze-issues

  Triage open GitHub issues — categorize, label, flag actionable ones, suggest closures with confirmation, and offer backlog creation via /kiln:kiln-report-issue.

• kiln-analyze

  Perform cross-artifact consistency analysis across spec.md, plan.md,

• kiln-build-prd

  Run the complete kiln pipeline using an agent team. Reads the PRD to determine team structure, then orchestrates specify → plan → tasks → implement → audit → PR.

• kiln-checklist

  Generate custom quality checklists for validating requirements completeness

• kiln-clarify

  Structured clarification workflow for underspecified requirements. Use

• kiln-claude-audit

  Audit CLAUDE.md against the kiln usefulness rubric and propose a git-diff-shaped removal/archival diff. Runs the full rubric (cheap + editorial signals). Never applies edits — output goes to .kiln/logs/ for human review.

• kiln-cleanup

  Remove stale QA artifacts from .kiln/qa/. Supports --dry-run for preview.

• kiln-constitution

  Create or update project governing principles and development guidelines.

• kiln-coverage

  Check test coverage against the 80% constitutional gate

• kiln-create-prd

  Create PRDs from product ideas or feature requests. Asks clarifying questions, then generates structured PRD artifacts. Supports three modes — PRD-only repo (multi-product), new product, and feature addition. Use when the user says "create a PRD", "PRD this", "add a feature", or similar.

• kiln-distill

  Bundle open backlog items from `.kiln/issues/`, open strategic feedback from `.kiln/feedback/`, AND structured roadmap items from `.kiln/roadmap/items/` into a feature PRD. Groups items by theme (feedback + current-phase items lead the narrative; issues are the tactical layer). Supports `--phase <na

• kiln-doctor

  Validate project structure against the kiln manifest and migrate legacy paths. Diagnose mode reports issues; fix mode applies corrections with confirmation. Supports cleanup of stale artifacts and version-sync checking.

• kiln-escalation-audit

  Inventory pause events from .wheel/history/, git log (confirm-never-silent), and .kiln/logs/ over the last 30 days. Emits a markdown report at .kiln/logs/escalation-audit-<timestamp>.md. V1 inventory-only — no verdict tagging. Use as /kiln:kiln-escalation-audit.

• kiln-feedback

  Log strategic product feedback about the core mission, scope, or direction to `.kiln/feedback/`. Distinct from `/kiln:kiln-report-issue` which captures bugs and friction. Use as `/kiln:kiln-feedback <description>`.

• kiln-fix

  Fix a bug without creating a new PRD or spec. Describe the issue (or pass a GitHub issue number) and the debugger will diagnose, fix, and verify it using the existing spec as context.

• kiln-hygiene

  Audit the repo against the structural-hygiene rubric and propose a review preview at .kiln/logs/. Never applies edits.

• kiln-init

  Initialize an existing repo with kiln. Installs the plugin, scaffolds project structure, and configures hooks — without creating a new GitHub repo. For repos that already exist.

• kiln-merge-pr

  Merge a PR and atomically flip its derived_from roadmap items in one operation. Wraps `gh pr merge` with a mergeability gate, post-merge state confirmation, PRD location, and an invocation of the shared `auto-flip-on-merge.sh` helper. Closes the async-merge auto-flip gap that PR #186 / PR #189 hit.

• kiln-mistake

  Capture an AI-made mistake (wrong assumption, bad tool call, missed context) as a schema-conformant note in .kiln/mistakes/. Shelf picks it up on the next sync and files a proposal in @inbox/open/ for human review. Use as "/kiln:kiln-mistake <free-form description>" or "/kiln:kiln-mistake" (skill wi

• kiln-next

  Analyze full project state and produce prioritized next steps mapped to kiln commands. Replaces /kiln:kiln-resume. Run at session start or after /kiln:kiln-build-prd.

• kiln-pi-apply

  Read open GitHub retrospective issues, extract File/Current/Proposed/Why PI blocks, and emit a propose-don't-apply diff report at .kiln/logs/pi-apply-<timestamp>.md. Never writes to skill/agent files — same discipline as /kiln:kiln-claude-audit and /kiln:kiln-hygiene. Use this to close the retro → s

• kiln-qa-audit

  Audit test files for duplicate scenarios and redundant assertions. Outputs a prioritized report to .kiln/qa/test-audit-report.md.

• kiln-qa-checkpoint

  Run a quick QA checkpoint — test recently completed flows, record video of failures, and send feedback to implementers. Fast iterative pass, not a full suite run.

• kiln-qa-final

  Quick E2E gate — run the Playwright test suite and report green/red. No evaluation, no issue filing, just confirm tests pass.

• kiln-qa-pass

  Standalone QA pass with 4-agent team. E2E tests + live /chrome testing + UX evaluation. Findings filed as GitHub issues. Use this outside the pipeline — for pipeline use, see /kiln:kiln-qa-pipeline.

• kiln-qa-pipeline

  Pipeline QA pass with 4-agent team. Same testing as /kiln:kiln-qa-pass but findings route to implementers via SendMessage for immediate fixing. Used by /kiln:kiln-build-prd, not standalone.

• kiln-qa-setup

  Install Playwright, scaffold QA test infrastructure, and generate a test matrix from the spec. Run this before any QA testing.

• kiln-report-issue

  Log a bug, friction point, or improvement idea to the project backlog. Creates a timestamped entry in .kiln/issues/. Use as "/kiln:kiln-report-issue <description>" or "/kiln:kiln-report-issue #42" to import from GitHub.

• kiln-research

  Run the baseline-vs-candidate research substrate against a declared fixture corpus. Emits a comparative markdown report at .kiln/logs/research-<uuid>.md with a strict-gate or per-axis-direction verdict (declared via PRD `empirical_quality:`). Required args: <baseline-plugin-dir> <candidate-plugin-di

• kiln-reset-prd

  Check out the base branch this feature branch was started from, abandoning the current feature branch.

• kiln-resume

  Deprecated — use /kiln:kiln-next instead. Runs /kiln:kiln-next with a deprecation notice.

• kiln-roadmap

  Capture a product-direction idea (feature / goal / research / constraint / non-goal / milestone / critique) as a structured item under `.kiln/roadmap/items/`. Runs an adversarial interview, classifies kind, routes cross-surface (issue / feedback / roadmap) with confirm-never-silent hand-off, migrate

• kiln-taskstoissues

  Convert tasks from tasks.md into GitHub issues. Use after task breakdown

• kiln-test

  Executable skill-test harness. Invokes real claude --print ... --plugin-dir subprocesses against /tmp/kiln-test-<uuid>/ fixtures, watched by a classifier that replaces hard timeouts. Usage: /kiln:kiln-test (auto-detect plugin), /kiln:kiln-test <plugin>, /kiln:kiln-test <plugin> <test>. Tests live un

• kiln-todo

  Quick task jotting without the full spec pipeline. Manage ad-hoc TODOs in .kiln/todos.md.

• kiln-ux-evaluate

  Standalone UI/UX evaluation using /chrome. Navigates every page, takes screenshots at each viewport, and evaluates heuristics, visual design, accessibility, and interaction quality. No functional testing — just design review.

• kiln-version

  Show the current version.

• plan

  Generate technical implementation plans from feature specifications.

• specify

  Create or update feature specifications from natural language descriptions.

• tasks

  Break down implementation plans into actionable task lists. Use after

Subagents (13)

• continuance

  Analyzes full project state and produces prioritized next steps mapped to kiln commands. Used by /kiln:kiln-next skill and as the final step in /kiln:kiln-build-prd.

• debugger

  Debug loop agent. Diagnoses and fixes bugs in already-implemented features without requiring a new spec or PRD. Classifies the issue, selects debugging techniques, runs a diagnose→fix→verify loop, and tracks failed approaches. Invoked directly by the user via /kiln:kiln-fix or spawned on-demand duri

• fixture-synthesizer

  Generates fixture corpora for empirical comparisons. Given runtime-injected role-instance variables (corpus shape, target count, axis constraints), produces deterministic fixture files that downstream research-runners measure against. Spawned by the research-first build-prd variant once per corpus.

• output-quality-judge

  Scores paired baseline/candidate outputs against a rubric. Given runtime-injected role-instance variables (rubric, output paths, axis weights), produces a structured judgment result the orchestrator uses to declare a comparison winner. Spawned by the research-first build-prd variant once per pairing

• prd-auditor

  Audits implementation compliance against PRD and feature specs. Checks BOTH directions: PRD→Spec (are all PRD requirements covered by FRs?) and Spec→Code→Test (are all FRs implemented and tested?). Fixes gaps or documents blockers.

• qa-engineer

  Visual QA engineer agent. Runs iteratively during implementation — tests user flows with Playwright, records video, and sends actionable feedback to implementers. Produces a final video report for the PR.

• qa-reporter

  QA reporting agent with two modes: 'issues' (files GitHub issues for standalone /kiln:kiln-qa-pass) and 'pipeline' (routes findings to implementers, waits for fixes, re-tests, then files remaining issues). Audits completeness in both modes.

• research-runner

  Reports empirical metrics for a baseline or candidate plugin against a fixture, given runtime-injected role-instance variables and verb-tool bindings. Spawned by the research-first build-prd variant in parallel — once per role-instance (baseline, candidate, ...) — with the same identity but differen

• smoke-tester

  Runtime smoke test agent. Acts like a human user — scaffolds a project in a temp dir, starts it, and verifies it actually works. Supports CLI, web (Playwright headless), and mobile (Maestro) project types.

• spec-enforcer

  Validates that code changes trace back to spec requirements

• test-runner

  Runs tests and reports results after code changes

• test-watcher

  Classifier agent for the /kiln:kiln-test harness. Polls a running test's scratch dir + NDJSON transcript and classifies session state (healthy / stalled / failed). Used by plugin-kiln/scripts/harness/watcher-runner.sh. Haiku-model — classification is cheap, generation is not the job.

• ux-evaluator

  UI/UX evaluation agent using 3-layer analysis: programmatic DOM checks (axe-core + custom JS), semantic accessibility tree review, and visual screenshot analysis. Sends all findings to qa-reporter.

Hooks (4)

• PreToolUse

  bash "${CLAUDE_PLUGIN_ROOT}/hooks/require-spec.sh"

• PreToolUse

  bash "${CLAUDE_PLUGIN_ROOT}/hooks/require-feature-branch.sh"

• PreToolUse

  bash "${CLAUDE_PLUGIN_ROOT}/hooks/version-increment.sh"

• PreToolUse

  bash "${CLAUDE_PLUGIN_ROOT}/hooks/block-env-commit.sh"

Similar plugins