| name | kiss-arch-extraction |
| description | Extracts architecture from an existing codebase — containers, components, data flow, external dependencies. Produces an architecture doc that can seed new ADRs and C4 diagrams. Use when reverse-engineering a codebase, onboarding to a new project, or when architecture documentation needs to be generated from existing code. |
| compatibility | Requires kiss project structure with .kiss/ directory |
| metadata | {"author":"github-kiss","source":"agent-skills/kiss-arch-extraction/kiss-arch-extraction.md"} |
| user-invocable | true |
| disable-model-invocation | false |
User Input
$ARGUMENTS
Audience and tone (interactive mode)
When KISS_AGENT_MODE=interactive (the default), assume the user
has limited technical background and limited domain knowledge
— they may know basics but lack deep expertise in this skill's
area. Run this skill as a guided questionnaire:
- One question at a time. No walls of questions.
- Yes / no first. Phrase so
yes, no, not sure, or skip
is a valid answer.
- Translate jargon, don't strip it. Use the technical term but
always pair it with a plain-English gloss the first time it
appears.
- Choices, not blank fields. When yes/no isn't enough, offer
2-4 lettered options (A/B/C/D) with one-line plain-language
descriptions of the trade-off. Always include "Not sure — pick
a sensible default".
- Always recommend. State the option you would pick and why in
one sentence so the user can reply "yes" / "ok" to accept. Pull
defaults from upstream artefacts (spec, architecture, ADRs,
standards) before asking blank.
- Show, don't ask. When upstream artefacts already imply an
answer, propose it as a pre-filled finding and ask for a
yes / no confirmation rather than asking the user to fill in a
blank.
not sure / skip triggers a sensible default, marked
"(default applied — confirm later)" in the artefact, and a debt
entry in this skill's debt file.
When KISS_AGENT_MODE=auto (or --auto), skip the questionnaire
entirely: apply sensible defaults from upstream artefacts and log
decisions to the parent agent's decision log.
Inputs
.kiss/context.yml
{context.paths.docs}/analysis/codebase-scan.md
- Source tree
Outputs
{context.paths.docs}/architecture/extracted.md
{context.paths.docs}/architecture/extracted.extract
Context Update
Does not mutate .kiss/context.yml.
Handoffs
kiss-c4-diagrams converts extracted structure into C4 Mermaid.
kiss-adr seeds candidate ADRs for decisions embedded in the
code (e.g. "chose PostgreSQL") that had no prior ADR.
AI authoring scope
Does: read config files (compose / k8s / terraform / ci),
entry-point modules, and infer the container + dependency graph.
Flag decisions that appear to have been made implicitly (no ADR).
Does not: modify code; claim intent without evidence.
Usage
<SKILL_DIR> = the integration's skills root (e.g. .claude/skills/
for Claude Code, .agents/skills/ for Antigravity / Codex,
.cursor/skills/ for Cursor, .windsurf/workflows/ for Windsurf).
Scripts live at <SKILL_DIR>/<skill-name>/scripts/….
bash <SKILL_DIR>/kiss-arch-extraction/scripts/bash/extract.sh --auto