Create a new capability for Amazon Bedrock

[johnfelipe]

Summary

I want to create a new folder, similar to Codex, Gemini, or Claude Code, but now for Amazon Bedrock.

Problem / use case

I already have a minimally coded process to perform the Amazon Bedrock process

GitHub copilot comes very close, but inputs must be enclosed in double curly braces, and a lot of syntax needs to be edited for all .md files and templates.

Proposed solution

Exact steps to create this new folder

Alternatives considered

No response

AI Tool

None

Additional context

No response

[johnfelipe]

pls review this attempt

main...johnfelipe:arc-kit:devin/1776169775-add-amazon-bedrock-plugin

but i dont know if skip something, prompt for amazon bedrock is with double {{}} and for pathing too double {{}}

[johnfelipe]

What do you think of my approach?

[tractorjuice]

Hi @johnfelipe — thanks for putting the time in (and Devin for the scaffolding). Reviewed the branch.

Happy to accept this as a community-maintained format, same model we already use for the EU/FR commands (@thomas-jardinet) and the upcoming AT seed (PR #320, awaiting @gtonic). That keeps the converter pattern clean and gives Bedrock users a first-class distribution without me having to track AWS-specific changes solo.

What I need from you to accept

1. CODEOWNERS commitment — willing to be auto-requested on any PR touching arckit-bedrock/*? Lightweight, no enforcement, you can step away anytime. I remain final approver on every PR. If yes, I'll add a @johnfelipe @tractorjuice line to .github/CODEOWNERS.

2. Open a real PR rebased onto current main. The branch is 52 commits behind — baseline has moved from 68 → 87 commands (7 EU + 11 FR + 1 IRN community-contributed), with 3 more AT commands queued in PR feat: add 3 Austrian government commands ([COMMUNITY] seed, inviting @gtonic as maintainer) #320.

3. Re-run python scripts/converter.py after the rebase so arckit-bedrock/prompts/ regenerates against the 87-command baseline (currently shows the stale 68).

4. Provenance markers — see how PRs feat: add 7 EU regulatory commands ([COMMUNITY], maintained by @thomas-jardinet) #314 / feat: add 11 French government commands ([COMMUNITY], maintained by @thomas-jardinet) #316 / feat(commands): add /arckit.fr-irn — IRN Indice de Résilience Numérique [COMMUNITY] #322 mark community work:

   • [COMMUNITY] prefix on the format-level files (your case: arckit-bedrock/README.md if you add one, plus the converter docstring for the amazon_bedrock entry)
   • Distinct from individual commands which already carry their own provenance

5. Tests — add a smoke test under tests/ mirroring the Paperclip pattern (tests/paperclip/test_commands_json.py) — at minimum a count assertion that prevents future converter regressions silently dropping Bedrock outputs.

6. Documentation — short arckit-bedrock/README.md covering: how a developer consumes the prompts in Bedrock Prompt Management, how the Step Functions JSON is intended to be deployed, and an explicit [COMMUNITY] notice that this format is community-maintained.

7. README / CHANGELOG entries — add Bedrock to the "Distribution formats" table in the top-level README.md, and a ### Added (Community-contributed) line in CHANGELOG.md under [Unreleased].

Strategic note (worth being explicit about)

The 6 existing formats all target AI coding assistants where a user types /arckit.requirements interactively. Bedrock is different — it's infrastructure for teams building agents, not an end-user CLI. That's a meaningful shift in audience for ArcKit, and it's why I want this clearly community-flagged: anyone using arckit-bedrock/ should know the prompts went through ArcKit's authoring discipline but the Bedrock-specific wiring (variable substitution, Step Functions orchestration) is community territory.

The {{var}} substitution and Step Functions JSON look like reasonable shapes — let's see them rebased before deep-reviewing the specifics.

Ready when you are.