feat: add GitHub Copilot instructions for contributors

[ChibaYuki347]

Summary

Add .github/copilot-instructions.md so contributors who use GitHub Copilot
to work on this repository get the same project context that Claude Code
(.claude/) and Kiro (.kiro/steering/) already receive. Content is a concise
subset of AGENTS.md (build/test/lint commands, code style, PR rules, security
scanners, and the do-not-rename public-contract constraints).

Scope note: This is intentionally limited to contributor tooling for the
current repository. It does not expand the product's supported-platform
surface. See the comment below for how this aligns with the v2 direction.

Changes

• .github/copilot-instructions.md — repository-wide Copilot context for contributors.

User experience

• Before: Copilot users editing this repo received no repo-specific guidance,
  unlike Claude Code and Kiro contributors.
• After: Copilot automatically loads the repo's conventions (commands, style,
  PR rules, constraints) from .github/copilot-instructions.md.

Checklist

• I have reviewed the contributing guidelines
• I have performed a self-review of this change
• Changes have been tested
• Changes are documented

Test Plan

• npx markdownlint-cli2 ".github/copilot-instructions.md" passes (0 errors).
• No .gitignore change needed — .github/** is version-controlled.
• No product files, README, or aidlc-rules/ content touched.

Acknowledgment

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of the project license.

[Copilot]

Pull request overview

Adds first-class GitHub Copilot contributor support to this repository by introducing repo-wide and path-specific Copilot instruction/prompt files, and documenting how AI-DLC workflows map onto Copilot surfaces (Ask/Plan/Agent, CLI, cloud agent). This complements the existing contributor configuration for other agents (e.g., .claude/, .kiro/steering/) and improves discoverability in existing docs.

Changes:

• Add repository-wide Copilot instructions plus path-scoped instruction files for aidlc-rules/**, scripts/aidlc-evaluator/**, and docs/**.
• Add reusable Copilot prompt files (/new-rule, /security-finding-triage) to standardize common contributor workflows.
• Add a dedicated Copilot integration guide and link it from existing contributor documentation and the README.

Show a summary per file

File	Description
README.md	Expands Copilot setup guidance, including recommended surfaces/modes and links to the new integration guide.
docs/GITHUB-COPILOT-INTEGRATION.md	New guide describing Copilot customization mechanisms and surface compatibility with AI-DLC runtime behavior.
AGENTS.md	Adds the new Copilot integration guide to the documentation map.
.github/copilot-instructions.md	New always-on Copilot context distilled for repo contributors.
.github/instructions/aidlc-rules.instructions.md	Adds path-scoped guidance for contributors editing the distributable aidlc-rules/** content.
.github/instructions/scripts.instructions.md	Adds path-scoped guidance for the Python evaluator subtree (uv/pytest + scanner conventions).
.github/instructions/docs.instructions.md	Adds path-scoped guidance for editing human-facing docs under docs/.
.github/prompts/new-rule.prompt.md	New reusable prompt to scaffold rule detail files under aidlc-rules/.
.github/prompts/security-finding-triage.prompt.md	New reusable prompt to triage scanner findings and apply fixes/suppressions correctly.

Review details

Tip

Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

• Files reviewed: 9/9 changed files
• Comments generated: 3
• Review effort level: Low

[Copilot]

Review details

• Files reviewed: 9/9 changed files
• Comments generated: 0 new
• Review effort level: Low

[ChibaYuki347]

Rescoped to avoid conflicting with the v2 direction

After reviewing the repository's direction, I've significantly reduced the scope of this PR.

What I found:

• main is still v1 (v1.0.1), with a "2.0 (Preview)" banner pointing to the v2 branch.
• v2 (RFC [RFC]: Restructure AI-DLC workflows #105) re-architects AI-DLC from a rules-only model to rules + skills + subagents + hooks running on a deterministic TypeScript/Bun engine ("one harness-neutral core rendered onto the CLI").
• The v2 harness set is Claude Code, Kiro IDE, Kiro CLI, and Codex CLI only — there is no harness/copilot/ anywhere in the v2 tree. GitHub Copilot (along with Amazon Q, Cursor, and Cline) is not carried over, because the v2 harness contract (manifest + hook adapter + emit.ts, per docs/harness-engineering/09-porting-to-a-new-harness.md) requires programmable hooks, subagents, and a local Bun runtime that Copilot's custom-instructions model does not provide.

What changed in this PR as a result:

• Removed the product-facing changes that would have promoted Copilot as a supported product platform (README GitHub Copilot section edits and a docs/GITHUB-COPILOT-INTEGRATION.md guide), since that runs counter to v2's direction.
• Removed the v1-structure-coupled files (.github/instructions/* and .github/prompts/*) that reference aidlc-rules/**, which would go stale once v2 lands.
• Kept only .github/copilot-instructions.md — contributor tooling for the current repo, orthogonal to the product's harness support and parallel to the existing .claude/ and .kiro/steering/ configs.

Happy to close this entirely if maintainers prefer, or to instead open a discussion about a proper v2 Copilot harness (and the hook/subagent primitives that would be prerequisites).

[ChibaYuki347]

Correction / addendum: Copilot's 2026 extensibility (my earlier comment was too pessimistic)

My previous comment stated that "the v2 harness contract requires programmable hooks, subagents, and a local Bun runtime that Copilot's custom-instructions model does not provide." After a deeper look at the current (2026) GitHub Copilot documentation, that claim is out of date and I want to correct it for the record.

Copilot now provides the primitives the v2 harness contract needs — most notably on the Copilot CLI (GA):

• Lifecycle hooks (sessionStart, postToolUse, preCompact, subagentStart/subagentStop, and agentStop — the analog of Claude Code's blocking Stop). Hook commands can run any local shell/script (e.g. bash: "bun run ./hooks/aidlc-adapter.ts"), and can block: preToolUse → permissionDecision: "deny" and agentStop/Stop → {"decision":"block","reason"} (with stop_hook_active to avoid loops). This is functionally equivalent to the settings.json/hooks.json + core-hook wiring the Claude and Codex harnesses use. https://docs.github.com/en/copilot/reference/hooks-reference
• Tool-restricted subagents with per-agent tools: allowlist and model: in .github/agents/*.agent.md, dispatched with isolated context (GA on CLI). https://docs.github.com/en/copilot/reference/custom-agents-configuration
• Agent Skills (.github/skills/<name>/SKILL.md, GA), local terminal execution, and MCP.

Status caveats: hooks are Preview in VS Code (and VS Code currently ignores hook matchers + uses different tool names, so an adapter shim must normalize them), and the cloud agent is a poor fit for AI-DLC's interactive approval gates (it runs autonomously to a PR, has an ephemeral FS, and doesn't ship bun).

Net: a harness/copilot/ for v2 is technically feasible — most cleanly targeting Copilot CLI, mapping closely onto the existing harness/codex/ pattern (a manifest.ts + an aidlc-copilot-adapter.ts hook shim + an emit.ts that generates the hook config/skills/trust). The remaining real constraints are model parity (v2 is tuned to Claude Opus 4.8 on AWS Bedrock — the biggest open question), VS Code hook maturity, and the porting work itself.

This does not change this PR. #470 stays correctly scoped to contributor tooling (.github/copilot-instructions.md) and does not add Copilot as a v2 product harness. I'll open a separate [RFC] to discuss whether a v2 Copilot harness is desirable and to work through the model-parity question with the maintainers.