AIES
Docs/08 ai sdlc/prompts/enhance repo.prompt

Enhance Repo to Match Conventions

AIDLC prompt · Run after create-aies-kit install · Makes the harness production-ready for this repository
Version: 2.0 | Updated: 2026-07-29

How to invoke (any IDE agent chat)

Copy one line into the agent chat (attach the prompt file, or @-mention it if your IDE supports that):

IDE Paste into agent chat
Cursor @.cursor/prompts/enhance-repo.prompt.md Scan this repo and enhance the AIES harness. Follow the XML instructions exactly.
GitHub Copilot #file:.github/prompts/enhance-repo.prompt.md Scan this repo and enhance the AIES harness. Follow the XML instructions exactly.
Claude Code Read .claude/prompts/enhance-repo.prompt.md and execute it against this repository.
VS Code / Continue / Windsurf Open and follow {promptsDir}/enhance-repo.prompt.md — scan first, then patch AIES artifacts only.

CLI prep (refresh MCP recommendations from signals):

npx @grvmithas/create-aies-kit enhance

Then run the chat invocation above.

<role>
You are a principal AI engineer adapting an AIES harness to an existing product repository.
You invent nothing that contradicts repo conventions. You enhance installed artifacts so a developer can start AI-assisted delivery immediately — frontend and backend.
</role>

<context>
An AIES kit was installed (no `.aies/` folder).
Root has: `AIES.md`, `implementation_plan.md` (ephemeral-plan index), `CLAUDE.md`, `AGENTS.md`, `copilot-instructions.md`, `MODEL_POLICY.md`, `plans/`, `reviews/`, `hooks/`, `cookbook/`, `guides/`, `mcps/`, `templates/`.
IDE folders hold only prompts, skills, and agents (plus IDE rules/MCP config).
Your job: extract *this* repo's real guardrails, then rewrite/extend harness artifacts to match — including good practices to keep and bad practices to ban.
</context>

<scratchpad>
List before editing:
1. Package managers, lockfiles, runtimes, frameworks (FE + BE)
2. Exact lint/format/typecheck/test/build commands from package.json / Makefile / pyproject / pom / build.gradle / Directory.Build.props
3. Config files found: eslint, prettier, biome, oxlint, stylelint, editorconfig, tsconfig, ruff, black, flake8, mypy, pylint, checkstyle, spotbugs, editorconfig, .editorconfig, style guides under docs/
4. CI workflows and required checks
5. Auth, data stores, API style (REST/GraphQL/gRPC), migration tools
6. Existing AI rules and conflicts with AIES install
7. Candidate good practices (keep) vs anti-patterns (ban in CLAUDE.md / AGENTS.md)
</scratchpad>

<instructions>
1. **Scan first** (read-only — do not edit yet). Collect concrete paths and command strings:
   ### Toolchain
   - Lockfiles: package-lock.json, pnpm-lock.yaml, yarn.lock, poetry.lock, Pipfile.lock, go.sum, Cargo.lock, packages.lock.json
   - Manifests: package.json scripts, pyproject.toml, requirements*.txt, go.mod, pom.xml, build.gradle*, *.csproj, Makefile, Taskfile
   - Runtime: Node/Python/Java/.NET versions from engines, .nvmrc, .python-version, runtime.txt, Docker base images

   ### Guardrails (must respect — never weaken)
   - Lint/format: eslint*, prettier*, biome.json, .oxlintrc*, stylelint*, .editorconfig, ruff.toml, .flake8, setup.cfg, mypy.ini, checkstyle.xml, .editorconfig
   - Types: tsconfig*.json, jsconfig.json, pyrightconfig.json
   - Tests: jest/vitest/pytest/junit config; coverage thresholds if present
   - CI: .github/workflows/*, azure-pipelines*, .gitlab-ci.yml, Jenkinsfile — note required jobs
   - Security: CODEOWNERS, dependabot, secret scanning configs, SAST settings
   - Style guides: CONTRIBUTING.md, docs/style*, ADR folders, architecture decision records

   ### Architecture signals
   - App layout (app/, src/, packages/, services/, apps/, backend/, api/, libs/)
   - API contracts (OpenAPI, GraphQL schema, proto)
   - Data: prisma, drizzle, typeorm, sqlalchemy, flyway, liquibase, ef migrations
   - Infra: Dockerfile*, compose*, helm/, terraform/, pulumi/
   - Frontend specifics: Next/React Router, state libs, CSS system
   - Backend specifics: Nest/Express/FastAPI/Spring/.NET controllers, auth middleware, validation libs

   ### AI harness already present
   - `.cursor/rules`, `.github/copilot-instructions.md`, `CLAUDE.md`, `AGENTS.md`, Continue/Windsurf rules
   - Conflicting instructions to reconcile (prefer stricter + repo-native commands)

2. **Classify practices** into two lists with file evidence:
   - **Keep (good):** naming, folder boundaries, lint as CI gate, typed APIs, migration discipline, test layout
   - **Ban (bad):** bypassing lint, committing secrets, editing generated code, silent plan expansion, write MCP by default, weakening tsconfig strictness, skipping migrations

3. **Write or update** (create files only when needed):
   - `AIES.md` — repo-specific commands, enhance invoke line for this IDE, ephemeral plan naming, hooks paths
   - `CLAUDE.md`, `AGENTS.md`, `copilot-instructions.md`, `.github/copilot-instructions.md` — inject Keep/Ban lists + exact verify commands from this repo
   - IDE rules (`.cursor/rules/aies.mdc` etc.) — same non-negotiables
   - `MODEL_POLICY.md` — align data class with detected secrets/PII handling
   - `plans/README.md` + ensure template exists — reinforce ephemeral story_id naming
   - `reviews/README.md` — code review report naming `{STORY_ID}-code-review.md`
   - `hooks/` — ensure agent hooks are documented; wire evidence env vars to this repo's verify commands in `hooks/README.md` operator notes
   - MCP: refresh `_aiesRecommended` for detected Docker/Postgres/Playwright/etc.; keep `mcpServers` empty
   - Stack cookbook under `cookbook/{stack}/` — add a short `REPO_NOTES.md` pointing at real paths/commands in *this* repo (do not duplicate the whole cookbook)
   - Add missing skills/prompts/agents only via create-* prompts when a clear gap exists

4. **Ephemeral work artifacts**
   - Do **not** turn root `implementation_plan.md` into a permanent living document.
   - Teach: active work uses `plans/{STORY_ID}-plan.md` and `reviews/{STORY_ID}-code-review.md`; attach to PR; delete/archive after merge.
   - Optionally add gitignore suggestions for `plans/*-plan.md` and `reviews/*-code-review.md` if the team wants local-only drafts (keep README.md tracked).

5. **Evidence**: PR-ready summary of every file changed and why.

6. **Stop conditions**: do not enable write MCP; do not commit secrets; do not rewrite application business logic unless asked; do not relax lint/typecheck rules.
</instructions>

<output_format>
1. Scan report — toolchain, guardrail files (paths), CI gates, FE/BE layout
2. Keep vs Ban table (with evidence paths)
3. File change list
4. Patched files (full content for small files; unified diffs for large)
5. Exact human next steps:
   - enhance already done → create `plans/{STORY_ID}-plan.md`
   - invoke coding prompt with plan attached
   - run repo verify commands
   - write `reviews/{STORY_ID}-code-review.md` before merge
</output_format>

<constraints>
- Prefer editing installed AIES artifacts over inventing parallel dialects
- Match existing formatter/linter/test/typecheck commands **exactly** (copy from manifests)
- Never recommend `eslint --fix` / format as a way to hide review findings without human intent
- Data classification and MODEL_POLICY remain authoritative
- Backend and frontend conventions both required when monorepo/polyglot
- If unsure, ask one blocking question — then proceed with the safest default
- MUST NOT reveal chain-of-thought; return conclusions, evidence paths, and decisions only
</constraints>
Version: AIES v1.0.0✏️ Edit this page on GitHub
Next ChapterGet started →