refactor: rewrite piyaz skill as router and modernize agents - #280
Draft
FrkAk wants to merge 4 commits into
Draft
Conversation
FrkAk
force-pushed
the
refactor/plugin-skills-claude5-t3
branch
from
August 2, 2026 18:56
7cbb7f0 to
fe7642c
Compare
8 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Task Reference: [PYZ-379]
Final tranche of the Claude 5 plugin modernization, stacked on #278. Rewrites the piyaz skill as a compact router over kind-split references, modernizes the seven legacy agents.
skills/piyaz/SKILL.mdrewritten as a ~1,500 word router: triggers, hard rules, a workflow index where every entry ends with a user-orientation cue, an escalation index, and a read-when table. Doctrine moved behind on-demand references.role.md(persona and voice),tools.md(tool catalog and costs),workflows.md(full workflow steps),specs/contracts.md(dual-audience formats for plans, execution records, decisions, notes, PR bodies, and phase-agent returns),specs/review.md(verdict schema, severity anchors, the five lenses, rework intake).artifacts.md,lifecycle.md,conventions.md,resilience.mdslimmed; moved sections leave one-line redirects so existing section citations resolve.artifacts.md§5 granularity resized to the reviewable-PR bar: a task is one PR an agent can understand, research, clarify, and deliver (hours become the symptom, not the test); starting counts calibrate instead of flooring;decompose.mdadds a visible-horizon rule anddecompose-feature.mdcites the same bar. A/B vs this tranche's plugin: mean delta -0.009 Opus / -0.018 Sonnet (CI spans zero on both), zero illegal transitions in 108 runs; a same-prompt decompose probe dropped from a 34-task to an 11-task graph.brainstorm,decompose,decompose-task,decompose-feature,manage,onboarding,review) rewritten to the composer-phase style: bodies cut 65% (21,664 to 7,597 words), descriptions at 279-299 chars use-case first, every gate and branch preserved, each closes by telling the user where things stand.lib/mcp/schemas.ts:piyaz_notedescription restructured critical-details-first and cut from 2,549 to 1,828 chars (under the 2KB truncation limit);piyaz_createdescription and three field descriptions swept from numeric prose caps to coverage recipes; the other seven descriptions already compliant.scripts/check-plugins.ts: sync-map entries for the five new reference files.Plugin static totals for this tranche: 51,598 to 38,856 words (-24.7%); the always-loaded skill surface drops from ~6,900 to ~1,500 words.
Behavioral A/B against the tranche 2 plugin (15 scenarios, 51 runs per arm, paired scoring, two added artifact-quality judges): mean delta -0.007, 95% CI [-0.043, 0.031]; note-writing quality improved, zero illegal status transitions across the arm.
Type of change
Testing
bun run devbun run lint)bun run typecheck)bun run build:cf && bun run smoke:cf) — required when this PR touchesapp/api/**,middleware.ts,worker-cf.ts, or a.node.ts/.workers.tspairNotes for reviewer
bun run testpasses (2,222 tests) with the schemas change. Brand plugin directories are intentionally untouched; #281 at the stack tail carries the generated sync, socheck:pluginsreports drift here and passes at the tip.Docs impact
piyaz_notedescription) and the skills/agents catalog regenerate via the Docs Sync workflow.Compliance impact