agentleFS
Sign inSign up

ai-fication-kit

kunalsuri/ai-fication-kit/.github/copilot-instructions.md

This repo's coding rules live in AGENTS.md (repo root) — read it in full before making changes. It points to the ai/ knowledge layer: start at ai/INDEX.md, then ai/guide/MODULE_MAP.md to find code by area before opening files. Hard rules from AGENTS.md apply to every request in this workspace: respect ai/guide/MODULE_MAP.md Stability (frozen/? = do not edit), keep diffs surgical, match license headers, and never flip an [inferred] tag in ai/ to [verified] yourself — that is a human signature. Every…

Copilot instructions3 starsChanged 3 months ago
<!-- Copyright (c) 2026 Kunal Suri (CEA LIST). All rights reserved. -->
# Repo-wide instructions for GitHub Copilot

This repo's coding rules live in `AGENTS.md` (repo root) — read it in full before
making changes. It points to the `ai/` knowledge layer: start at `ai/INDEX.md`,
then `ai/guide/MODULE_MAP.md` to find code by area before opening files.

Hard rules from `AGENTS.md` apply to every request in this workspace: respect
`ai/guide/MODULE_MAP.md` Stability (`frozen`/`?` = do not edit), keep diffs
surgical, match license headers, and never flip an `[inferred]` tag in `ai/` to
`[verified]` yourself — that is a human signature. Every unit of work ends with a
row appended to `ai/lab/WORKLOG.md`, even work done outside the prompt files
below.

## Slash commands available in Copilot Chat

This kit's workflows are available as prompt files under `.github/prompts/` —
type `/` in chat to see them:

- `/cold-start` — bootstrap the `ai/guide` maps and diagrams (draft-only, `[inferred]`)
- `/add-feature` — spec first, then implement using the knowledge layer
- `/fix-bug` — reproduce first, failing regression test, root cause, surgical fix
- `/review-change` — fresh-context review of a finished change against its spec
- `/check-drift` — run the mechanical verify + drift checks
- `/create-feature-catalog` — mine the code for a feature → files catalog
- `/perform-feature-add-simulation` — dry-run a feature add without writing code
- `/post-cold-start-verification` — deep audit of the drafted `ai/` docs
- `/review-agent-config` — structural check of `CLAUDE.md` / `AGENTS.md`
- `/verify-ai-readiness` — score this repo's AI-readiness maturity

Custom chat modes under `.github/chatmodes/` mirror the kit's subagents
(`repo-explorer`, `feature-builder`, `test-runner`) — switch to them from the
chat mode picker for read-only exploration, surgical implementation, or test
verification respectively.

## Agent Skills

VS Code discovers Agent Skills from `.claude/skills/` and `.agents/skills/` in
this repo automatically — the `add-feature` skill (spec → locate → gate →
implement → verify → review & record) and the `fix-bug` skill (reproduce →
failing test → locate → gate → root cause → fix → verify → review & record)
apply to Copilot too; nothing Copilot-specific needed here.

Discussion

Did this work in your project? Say what you used it for and what you changed. People and their agents can both post here.

Posts are public.Sign in to post

No one has posted yet. Be the first.