AgentSkillsCN

repo-contract-alignment

通过扫描合约文档、提取明确规则、执行各类检查，并以强制执行的方式实施最小化的合规性修复，使仓库与其所宣称的治理方针保持一致。

SKILL.md
--- frontmatter
name: repo-contract-alignment
description: "Align the repository with its stated governance by scanning contract docs, extracting explicit rules, running checks, and implementing the smallest compliant fixes with enforcement."
<!-- CONTENT_HASH: dc5da4881546efe848cbadf457ccc141e4512024375fae7f5649fa98839fc310 -->

required_canon_version: >=3.0.0

Skill: repo-contract-alignment

Version: 0.1.0 Status: Active

Purpose

Align the repository with its stated governance by scanning contract docs, extracting explicit rules, running checks, and implementing the smallest compliant fixes with enforcement.

Trigger

Use when asked to align repo behavior with canon/contract docs, audit governance compliance, or implement minimal fixes plus enforcement for regressions.

Workflow

  1. •

    Confirm intent gate

    • •If the user is asking for analysis or strategy only, stop after Findings/Plan and ask for explicit implementation approval.
    • •Do not touch CANON or edit existing CONTEXT unless the task is explicitly about rules/governance/memory updates.
  2. •

    Ensure cortex is available

    • •If CORTEX/_generated/cortex.db is missing, run python CORTEX/cortex.build.py.
    • •Use python CORTEX/query.py --json or --find for file discovery; avoid raw filesystem scans.
  3. •

    Identify contract docs

    • •From the cortex index, collect docs matching: README.md, AGENTS.md, CANON/, CONTEXT/decisions/ (ADRs), ROADMAP, CONTRIBUTING, SECURITY, CONTEXT/maps/*, and any explicit “contract” docs.
  4. •

    Extract explicit rules

    • •Read only those contract docs and list their rules.
    • •Resolve conflicts using the authority gradient; note any conflicts.
  5. •

    Run existing checks/tests/build scripts

    • •Always run python TOOLS/critic.py and python CONTRACTS/runner.py.
    • •Run any additional scripts referenced by contract docs.
  6. •

    Report issues and minimal fixes

    • •Produce a prioritized issue list (P0 violations, P1 missing enforcement, P2 hygiene).
    • •Propose the smallest fixes and enforcement to prevent regressions.
  7. •

    Implement and re-check

    • •Apply the smallest correct changes, add/adjust fixtures if behavior changes, then re-run checks until passing.
    • •Write artifacts only to allowed roots (CONTRACTS/_runs/, CORTEX/_generated/, MEMORY/LLM_PACKER/_packs/).

Output format

Always respond with: Findings, Plan, Changes, Commands run, Next.

required_canon_version: >=3.0.0