Update

You are updating the vendored Sprint chain in already-bootstrapped repos from the freshly-installed plugin templates. The plugin ships the canonical agents/ + .mcp.json (those auto-update with the plugin); this command refreshes the vendored copies that the headless git-hook path depends on, plus the stage scripts and hooks.

Only refresh logic. Never touch the user's config or content: leave sprint.conf, TODO.md, DONE.md, AGENT_LOG.md, AGENTS.md, CLAUDE.md and .gitignore alone.

0. Locate templates + the target repo(s)

Templates come from the plugin as installed — never from a development checkout:

TPL="${CLAUDE_PLUGIN_ROOT:-}/templates"
# Fallback when CLAUDE_PLUGIN_ROOT is unset (e.g. skill executed manually): resolve the
# INSTALLED plugin from the marketplace cache — most recently installed version wins.
[ -d "$TPL" ] || TPL="$(ls -dt "$HOME"/.claude/plugins/cache/*/sprint/*/templates 2>/dev/null | head -1)"
[ -d "${TPL:-}" ] || { echo "sprint is not installed. Run:
  claude plugin marketplace add leonhoffmann86/team-sprint
  claude plugin install sprint@team-sprint"; }
REG="${XDG_CONFIG_HOME:-$HOME/.config}/sprint/registry"   # one repo path per line

If neither resolves, stop with that install instruction. Do NOT search the filesystem for a plugin source tree and do NOT accept a git checkout of the plugin repo as $TPL — only the installed (versioned) copy is a valid template source (see docs/DISTRIBUTION.md).

• No argument → operate on the current repo only: ROOT="$(git rev-parse --show-toplevel)".
• --all → read $REG (if present) and operate on each listed path that is still a git repo with a scripts/sprint-lib.sh (i.e. actually bootstrapped). Report missing/!bootstrapped entries and skip them. If $REG is absent, tell the user no repos are registered and that /sprint:bootstrap records repos there (see step 4).

1. Refresh the logic files (overwrite — these are NOT user-owned)

For each target ROOT, copy over (overwrite) ONLY:

cp "$TPL/scripts/sprint-lib.sh"       "$ROOT/scripts/sprint-lib.sh"
cp "$TPL/scripts/sprint-plan.sh"      "$ROOT/scripts/sprint-plan.sh"
cp "$TPL/scripts/sprint-implement.sh" "$ROOT/scripts/sprint-implement.sh"
cp "$TPL/scripts/sprint-review.sh"    "$ROOT/scripts/sprint-review.sh"
cp "$TPL/scripts/sprint-gate.sh"      "$ROOT/scripts/sprint-gate.sh"
cp "$TPL/githooks/post-commit"        "$ROOT/.githooks/post-commit"
cp "$TPL/githooks/README.md"          "$ROOT/.githooks/README.md"
mkdir -p "$ROOT/.claude/agents"
cp "$TPL"/.claude/agents/*.md         "$ROOT/.claude/agents/"
chmod +x "$ROOT/.githooks/post-commit" "$ROOT"/scripts/sprint-*.sh

For .mcp.json: if the repo's file is identical to a prior template version, overwrite it; if the user has customized it (extra servers), merge the codegraph server in rather than clobbering — show the diff and confirm. Before overwriting any file that the user may have hand-edited, show a diff and ask.

2. Surface config drift (don't auto-edit sprint.conf)

Compare the keys defined in $TPL/sprint.conf against the repo's sprint.conf. For any key present in the template but missing in the repo (e.g. a newly added SPRINT_*), print it with its default and a one-line description, and tell the user to add it if they want the new behavior. Never rewrite their sprint.conf automatically.

3. Sanity-check

Run bash -n on each refreshed script; report any failure.

4. Tooling report (MANDATORY — the chain lives on its tool use)

The chain degrades gracefully when supporting tools are missing, but the user must be told EXPLICITLY. Check and report each:

• codegraph (https://github.com/colbymchenry/codegraph): command -v codegraph and the repo index .codegraph/codegraph.db. Missing binary → roles run WITHOUT code-graph intelligence; missing index → suggest codegraph sync . once.
• fallow (https://docs.fallow.tools): on PATH or ./node_modules/.bin/fallow (unless SPRINT_FALLOW=off). Missing → the gate runs without static analysis; suggest npm i -g fallow + fallow init.
• jq and timeout/gtimeout: missing → degraded JSON parsing / no per-phase timeout. For anything missing, print the install command and the concrete impact — the same status appears as the ### Tooling section in every TODO.review.md.

5. Registry (consume-only — NEVER self-register)

$REG is a newline-delimited path list consumed by --all. Registration is exclusively /sprint:bootstrap's job and opt-in there — this command must NOT add the current repo (users keep internal repos out of the registry deliberately; see docs/DISTRIBUTION.md). If an entry no longer exists or isn't bootstrapped, drop it and note the cleanup.

6. Summarize

Print, per repo: which files were refreshed/skipped, any .mcp.json merge, the config-drift keys to consider, the tooling report, and the reminder that sprint.conf + lifecycle files were intentionally left untouched.