本文へ移動
cccskills
無料GitHub で公開

airlineops-parity-scan

Use to mine ../airlineops for governance habits gzkit lacks; produces a parity scan report (gzkit leads, AirlineOps adopts).

インストール方法を見る

含まれるファイル(1)

  • SKILL.md8.9 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

SKILL.md

AirlineOps Parity Scan

Run a repeatable governance parity scan that mines ../airlineops for process habits gzkit lacks.

Direction (AGENTS.md § Architectural Boundaries, boundary 5): gzkit leads and AirlineOps adopts. AirlineOps is part of gzkit's genesis and its first adopter, not a canon gzkit must match. A Missing or Divergent row is a candidate gzkit weighs by the intake rubric, never an obligation to match. Where this skill says "canonical" of AirlineOps or "extraction" of gzkit, it names the direction of the mining read and confers no authority.

Trigger

  • Weekly parity check cadence.
  • Before promoting any pool ADR into active implementation.
  • On the operator's request. A change in AirlineOps does not trigger a scan: gzkit does not chase AirlineOps patches.

Behavior

Produce a dated parity report with explicit gaps, severity, and required ADR/OBPI follow-up. Parity is operational, not textual. The scan MUST evaluate behavior/procedure sources across multiple surfaces.

Operational identity rule:

GovZero = AirlineOps - (AirlineOps product capabilities)

Treat every AirlineOps process habit/structure/rule as GovZero parity scope unless it is purely product capability behavior.

GovZero mining rule:

  • Treat AGENTS.md, governance docs, and control-surface directories (.github/, .claude/, .codex/, .gzkit/) as first-class GovZero extraction sources.
  • Mine process habits from all of them; do not restrict extraction to one docs subtree.

Prerequisites

  • ../airlineops exists and is readable.
  • gzkit repo is initialized with docs/ and docs/design/.
  • Use canonical references in docs/design/lodestar/govzero-doctrine.md.

Steps

  1. Confirm both repos are present: test -d ../airlineops && test -d .
  2. Resolve canonical root deterministically:
    • explicit override (if provided)
    • sibling path ../airlineops Fail closed if none resolve. There is deliberately no absolute fallback: a hardcoded machine path is a rung that can only ever resolve for one reader, which makes a ladder look deterministic while being shorter than it reads for everyone else (GHI #900). Pass an explicit override instead. Record canonical-root evidence in the parity report.
  3. Read canonical source declarations: docs/design/lodestar/govzero-doctrine.md, docs/design/lodestar/README.md, AGENTS.md.
  4. Build a behavior/procedure source matrix from canonical and extraction surfaces:
    • Canonical behavior sources (AirlineOps):
      • .claude/skills/gz-*
      • .github/skills/gz-* (where present)
      • .github/instructions/*.instructions.md (where present)
      • .claude/**
      • .codex/**
      • .gzkit/**
      • AGENTS.md, CLAUDE.md
      • docs/governance/GovZero/**/*.md
      • operator runbook/proof docs where rituals are declared
    • gzkit extraction surfaces:
      • .gzkit/skills/* (canonical) and the vendor mirrors named in .gzkit.json § vendors
      • .claude/**
      • .codex/** (if present)
      • .gzkit/**
      • AGENTS.md, CLAUDE.md
      • docs/governance/GovZero/**/*.md
      • docs/user/manpages/*, docs/user/concepts/*, docs/user/runbook.md
      • runtime control surfaces (src/gzkit/cli/, ledger events, validation commands)
  5. Produce a GovZero mining inventory:
    • list each mined norm/habit
    • cite source file/path
    • map to extracted gzkit surface (or mark missing/divergent)
    • classify confidence and remediation target
  6. Apply parity intake rubric to each candidate import:
    • Use docs/governance/parity-intake-rubric.md
    • Classify each item as Import Now, Import with Compatibility, Defer (Tracked), or Exclude
    • Record rationale and runtime/proof backing for the classification
  7. Compare both artifact parity and procedure parity:
    • artifact parity: presence/path/content
    • procedure parity: orientation, tool use, post-accounting, validation, verification, presentation, human authority boundaries
  8. Classify each matrix item as:
    • Parity
    • Partial
    • Missing
    • Divergent
  9. Execute and record runnable ritual checks from gzkit surfaces:
    • uv run gz cli audit
    • uv run gz check-config-paths
    • uv run gz adr audit-check ADR-<target>
    • uv run mkdocs build --strict
  10. Write a dated report: docs/proposals/REPORT-airlineops-parity-YYYY-MM-DD.md Use docs/proposals/REPORT-TEMPLATE-airlineops-parity.md. Also produce mining inventory report: docs/proposals/REPORT-airlineops-govzero-mining-YYYY-MM-DD.md Use docs/proposals/REPORT-TEMPLATE-airlineops-govzero-mining.md.
  11. For every Missing, Divergent, or high-impact Partial finding:
  • Route it by AGENTS.md § Defect-fix routing: name the owning ADR when the gap falls under existing intent, a GHI when it is a defect, and propose a new ADR only for a new capability
  • Add an explicit next action
  1. Summarize risk:
  • What blocks 1.0 readiness
  • What can wait
  • What must be done next cycle
  1. Enforce action closure for this scan cycle:
  • For each Import Now finding, execute at least one concrete tranche in the same cycle (code/docs/settings change) and record file-level evidence. A src/** change needs a GHI; OBPI-shaped work is proposed to the operator, not started.
  • If execution cannot happen in-cycle, file a defect through /ghi-author, whose prior-art lookup runs first, and link it in the report.
  • A scan with findings but no execution artifact or defect artifact is INCOMPLETE.

Examples

Example 1

Input: "Run the weekly AirlineOps parity scan."

Output: docs/proposals/REPORT-airlineops-parity-2026-02-07.md with matrix, severity, and action list.

Constraints

  • Do not modify anything in ../airlineops.
  • Do not claim parity without path-level evidence.
  • Do not claim procedure parity without executable ritual evidence from gzkit runtime/docs surfaces.
  • Do not use patch versions for new feature sequencing.
  • Keep findings actionable: each gap names its route (owning ADR, GHI, or proposed ADR).
  • Do not close a scan cycle with only passive tracking; require execution or a filed defect.

Common Rationalizations

These thoughts mean STOP — you are about to ship a passive-tracking parity report:

ThoughtReality
"The matrix is filled in — that closes the cycle"A scan with Missing or Divergent findings and no execution artifact (or filed defect) is INCOMPLETE by definition. Step 13 is mandatory.
"I'll execute the Import Now items next cycle"Action closure is per-cycle. Deferring is exactly the perpetual-catch-up pattern the constraint was written to prevent. Ship at least one tranche or file a defect now.
"Textual diff is good enough for parity"Parity is operational, not textual. You must mine procedural sources across all surfaces (.github/, .claude/, .codex/, .gzkit/, AGENTS.md) and exercise rituals from the gzkit runtime.
"I don't need to read AirlineOps docs — gzkit leads now"gzkit leads, and the scan still reads AirlineOps: the mining inventory is how gzkit finds a habit worth importing. Without it the report is a self-portrait, not a parity scan.
"Most items are Parity, the report is mostly done"The report is the easy part. The failing parity items are the work. Don't conflate report completeness with cycle completeness.
"I can write the report without running the runnable ritual checks"Step 9 ritual checks (gz cli audit, gz check-config-paths, gz adr audit-check, mkdocs --strict) provide the procedure-parity evidence. Without them you're claiming parity by inspection.
"Findings only need passive defer (tracked) classification"The constraint explicitly prohibits closing a cycle with only passive tracking. Each finding needs an execution path or a tracked defect — there is no third option.

Red Flags

  • Report contains Missing or Divergent findings but no commits, diffs, or filed GHIs in the same cycle
  • Mining inventory cites only one or two surface roots instead of the full set
  • Procedure parity claimed without runtime ritual evidence
  • The matrix lists items as Partial without specifying what's missing
  • Parity classified Defer (Tracked) for more than half the findings (perpetual-catch-up signal)
  • Report omits canonical-root resolution evidence (Step 2)
  • Cycle closed without producing both the parity report AND the mining inventory report

Related Skills

  • gz-adr-create
  • gz-adr-audit

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Use to file a GitHub issue when a finding needs its own work order or the operator asks; OBPI corrections stay in its change log.

日本語の概要は準備中です。原文の説明を表示しています。

tvproductions/gzkit92026年10月18日 更新

ghi-close

無料

Use to do the work a GHI describes, verify it and close the issue with evidence in one sweep.

日本語の概要は準備中です。原文の説明を表示しています。

tvproductions/gzkit92026年10月18日 更新

Use to triage the open GHI queue before planning or choosing what to pull; produces a ranked deliverable.

日本語の概要は準備中です。原文の説明を表示しています。

tvproductions/gzkit92026年10月18日 更新

git-sync

無料

Run the guarded repository sync ritual; commit-time hooks enforce the cheap gates and the pre-push hook runs gz check.

日本語の概要は準備中です。原文の説明を表示しています。

tvproductions/gzkit92026年10月18日 更新

git-sync

無料

Run the guarded repository sync ritual; the hooks your pre-commit config declares enforce the quality gates.

日本語の概要は準備中です。原文の説明を表示しています。

tvproductions/gzkit92026年10月18日 更新

Use for an ADR's Gate-5 audit; provides the audit templates and procedure.

日本語の概要は準備中です。原文の説明を表示しています。

tvproductions/gzkit92026年10月18日 更新

tvproductions のスキルをすべて見る

このスキルの問題を報告する