Skip to content
Back to skills

Repository Tooling Change Builder

ASecurity

Use `task-agent` for a bounded repository generator, plugin, harness, CLI, monorepo automation, or maintenance utility source change. Skip product behavior, config-only work, docs, release, review, and planning.

  • 7 stars
  • 0 votes
  • 0 copies
  • 4 views
  • Added September 8, 2026
toolsgoapibackenddocumentation

Works with

  • cli
  • api

Security analysis

A100/100

Pro scans all 4 files and shows the line behind each finding

Scanned September 22, 2026

npx -y skills add machenjie/rd-skills --skill repository-tooling-change-builder --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Repository Tooling Change Builder?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Repository Tooling Change Builder
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/machenjie-repository-tooling-change-builder/badge)](https://www.skillsdirectory.com/skills/machenjie-repository-tooling-change-builder)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: repository-tooling-change-builder
description: "Use `task-agent` for a bounded repository generator, plugin, harness, CLI, monorepo automation, or maintenance utility source change. Skip product behavior, config-only work, docs, release, review, and planning."
---

# repository-tooling-change-builder

## Role

Support `task-agent` in changing bounded repository tooling with authority, compatibility, determinism, cleanup, and proof.

## When To Use

- one bounded repository code generator, compiler plugin, linter, formatter, harness, internal CLI, build or monorepo automation, or maintenance utility change
- a direct task whose primary consumers are repository developers, tests, builds, or local automation

## Do Not Use

- backend service, product behavior, or business logic
- build-graph or build-configuration-only work with no repository-tool source change
- documentation-only work
- production mutation, deployment, release, or rollback
- independent review, task decomposition, planning, routing, or control

## Required Inputs

- behavior, scope, owner, consumer, invocation, tests, and outcomes
- source/generated authority, toolchain/host, effects, compatibility, cleanup, and rollback

## Professional Decision Rules

- Honor tooling ownership, inputs, stops and output contract.
- Bind generated output/plugins to authoritative inputs, source ownership, destination, tool/host versions, compatibility, deterministic identity and non-circular clean-checkout bootstrap.
- Bind mutating commands/subprocesses to resolved targets and invocation contracts, atomic completion or explicit recovery, cancellation/child cleanup and safe reruns.
- Prove harness oracles with valid/invalid controls; distinguish harness health from changed behavior correctness.

## High-Value Gotchas

- Generated output can look correct while bootstrap order or source authority is wrong.
- A subprocess can report success before output, cleanup, or child completion is durable.
- A maintenance command can cross its intended workspace or mutate files on rerun.

## Execution Checklist

- **Task mode:** Map source authority, callers, generated outputs, side effects, and cleanup ownership.
- Reuse the existing command, generator, harness, or process boundary when it owns the behavior.
- Verify compatibility, deterministic output, atomic completion, and bounded rerun behavior.
- Record skipped hosts, toolchains, consumers, and recovery paths as proof limits.
- Minimal validation: run normal, invalid, boundary, rerun, and forbidden-effect tests.

## Stop / Escalation Conditions

- Inspect affected owners, shared contracts, and dependencies directly. Request deeper analysis only when a concrete unresolved decision could change implementation.
- Stop on unresolved authority, bootstrap, compatibility, oracle, recovery, or validation.
- Keep other work at its owner; do not claim review.

## Output Contract

- changed files, owner, consumer, reuse, placement, invocation, compatibility, authority, bootstrap, file/process, and failure decisions
- validation, negative controls, skips, diff, cleanup/rollback, proof limits, residual risk, and boundary handoffs

## Targeted References

| Path | Type | Load when | Do not load when | Required by | Required output |
|---|---|---|---|---|---|
| [generator and plugin contracts](references/generator-and-plugin-contracts.md) | targeted | Generated authority, bootstrap, compiler protocol, host API, diagnostics, fixes, or version compatibility changes | No generator or compiler/linter/formatter plugin behavior changes | task-agent | boundary-decision, selected-approach, proof-limit |
| [harness validity contracts](references/harness-validity-contracts.md) | targeted | Test discovery, orchestration, oracle, fixtures, negative controls, benchmark measurement, or harness exit behavior changes | Existing unchanged harness directly proves the accepted behavior | task-agent | decision-record, validation-plan, proof-limit |
| [repository automation contracts](references/repository-automation-contracts.md) | targeted | Internal CLI, hook, monorepo automation, maintenance mutation, subprocess, rerun, or cleanup behavior changes | No repository automation interface or mutation changes | task-agent | decision-record, failure-decision, proof-limit |

Files in this skill

  • SKILL.md5.7 KB
  • references/generator-and-plugin-contracts.md3.1 KB
  • references/harness-validity-contracts.md2.9 KB
  • references/repository-automation-contracts.md3 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

View all in tools →

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.