Bootstrap GSD + graphify for an existing brownfield repo when the project needs a single canonical workflow for local runtime setup, graph refresh, and manually seeded .planning/ context without depending on interactive GSD init.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Bootstrap GSD + graphify for an existing brownfield repo when the project needs a single canonical workflow for local runtime setup, graph refresh, and manually seeded .planning/ context without depending on interactive GSD init.
Keep this explicitly labeled as fallback so teammates are not forced into a heavier path.
Brownfield Bootstrap Steps
1. Install local Codex runtime in the target repo
From the target project root:
npx -y get-shit-done-cc@latest --codex --local
If that entrypoint is unavailable, use the source-install fallback above.
Verify:
find ./.codex -maxdepth 3 -type f | sort | sed -n '1,120p'
2. Confirm graphify outputs and add smart refresh
Required outputs are:
graphify-out/graph.json
graphify-out/GRAPH_REPORT.md
Important rule:
do not require graphify-out/manifest.json
newer graphify versions may not emit it
treat manifest as optional diagnostic output only
graphify 0.5+ may create transient .graphify_chunk_*.json files during Codex semantic extraction; do not clean those before collection finishes
if the brownfield corpus includes video/audio, install graphifyy[video] or surface manual transcription guidance before expecting a complete semantic graph
The repo-local scripts/graphify-sync.sh should support:
status
smart
force
serve
Recommended smart behavior:
if graph outputs are missing, run graphify update .
if code files changed, run code-only rebuild via graphify.watch._rebuild_code(Path('.'))
if only docs/media changed, skip automatic rebuild and print manual full-refresh guidance
3. Add or verify repo docs and workflow contract
Update or create:
AGENTS.md
README.md
docs/current-state.md
docs/index.md
docs/entrypoints.md
optional docs/gsd-graphify-workflow.md
Converge them around current facts:
authoritative frontend and backend entrypoints
current mainline architecture path
what is legacy / experimental / archived
the standard GSD + graphify operating loop
4. Manually seed .planning/ when interactive init is not trustworthy
If the repo is complex, credentials are missing, or the current architecture already has a mature analysis baseline, do not block on gsd-sdk init.
act as the machine-readable task selector for autonomous continuation
map each major brownfield step to explicit task ids, dependencies, priorities, and acceptance notes
stay aligned with REQUIREMENTS.md, ROADMAP.md, and STATE.md
5. Optional cleanup of legacy entrypoints
For repos with multiple old entrypoints or migration leftovers:
document authoritative entrypoints first
archive low-risk legacy entrypoints before deleting anything
do not start by deleting the most ambiguous historical mainline files
This keeps the brownfield migration reversible while the new workflow settles in.
Recommended Phase 0
A good first brownfield phase is usually:
00-01 planning bootstrap
00-02 docs/current-state alignment
00-03 workflow verification
Only after that should the repo move into larger cleanup or architecture-convergence phases.
Verification Checklist
Run and inspect these before claiming success:
find ./.codex -maxdepth 3 -type f | sort | sed -n '1,120p'
gsd-sdk --version
gsd-sdk --help
./scripts/graphify-sync.sh status
./scripts/graphify-sync.sh smart
find .planning -maxdepth 3 -type f | sort
Also verify:
graphify-out/GRAPH_REPORT.md exists
graphify-out/graph.json exists
README / docs / AGENTS agree on current facts
.planning/ROADMAP.md and .planning/STATE.md point at the same active phase
if legacy entrypoints were archived, their replacements are clearly documented
Common Pitfalls
Treating source-install as the default teammate path
prefer the standard upstream entrypoints first
Requiring manifest.json
manifest is optional in current graphify versions
Letting gsd-sdk init invent a roadmap for a complex repo with weak context
manually seed .planning/ when needed
Updating scripts but not README / docs / AGENTS
brownfield repos need documentation convergence, not only tool installation
Deleting historical entrypoints before documenting the current mainline
convergence first, destructive cleanup second
Standard Operating Loop After Bootstrap
./scripts/graphify-sync.sh smart
Read graphify-out/GRAPH_REPORT.md
Read .planning/STATE.md and .planning/ROADMAP.md
Use GSD phase / plan / execute workflow
Implement or clean up
Re-run ./scripts/graphify-sync.sh smart
Update planning/docs if the phase meaning changed
Canonical Skill Note
This is the only brownfield bootstrap skill that should remain in the shared workflow set.
If you find references to brownfield-gsd-graphify-bootstrap, update them to this skill and remove the duplicate.
