From codepresso
Scaffold an AI-native AIDLC repo structure into a target path — analyze, interview (situational branching), preview, apply (missing only), re-score. Non-destructive, minimal external dependencies.
How this skill is triggered — by the user, by Claude, or both
Slash command
/codepresso:aidlc-initThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
<Purpose>
<Use_When>
/codepresso:aidlc-init <target-path> invokedAnalyze (CLI): run node "${CLAUDE_PLUGIN_ROOT}/scripts/aidlc-cli.mjs" detect <path> then ... scan <path>.
Show a compact 18-item table (status + evidence) and the overall %. If secrets[] is non-empty, surface a 🔴 warning at the top and tell the user to rotate them — never copy a secret value anywhere.
Interview (AskUserQuestion — 7 branches; only the non-inferable). Build a profile from the answers. Pre-fill each branch from detect:
detect.tools): which tools the team actually uses (claude, cursor, opencode, cline, copilot, gemini, amazonq). Only selected tools are scaffolded/scored; the rest become na.detect.host. If host is none, ask "create GitHub Actions anyway, or skip?" (default skip → ci-pr na).raw (a .git/hooks/pre-push wired by hand, zero deps). Offer husky/lefthook only if detect.hookFramework indicates it or the user asks. skip → na.regen-node (bundled zero-dep scanner, stays fresh). Opt-in: external tool (richer, adds a dependency) or static (one-time, will go stale) or off.detect.localDev shows a one-command bring-up, set localDev: 'detected' and inject that command into AGENTS.md. Otherwise offer localDev: 'scaffold' (apply-static writes an executable scripts/local-up.sh) or localDev: 'skip'.TSK) or "none".Write the answers to .codepresso/state/aidlc-profile.json:
{ tools[], ciHost, prePush, contextMode, localDev, integrations[], ticketPrefix }.
Preview (CLI, profile-aware): run ... scan <path> --profile .codepresso/state/aidlc-profile.json (honest na-scoped score) then ... plan <path> --profile <profile>. Show the file tree to be created, marking each static (templated) vs authored (skill-written). List any still-MISSING item NOT in the plan and not authored below (e.g. hooks, unit-tests, runbook) as 'manual setup required — not auto-created'. Ask for explicit confirmation. Write nothing before confirmation.
Apply:
a. Static: ... apply-static <path> --profile <profile> (templated, non-destructive; host-correct CI variant picked from ciHost).
b. Wire, don't just create:
pre-push: after the script exists, actually WIRE it — raw: copy/symlink to .git/hooks/pre-push + chmod +x; husky/lefthook: add the pre-push entry. A created-but-unwired script only scores partial.contextMode=regen-node: merge the SessionStart hook from templates/aidlc/.claude/settings.codesight-hook.json into the repo's .claude/settings.json, then run node "${CLAUDE_PLUGIN_ROOT}/scripts/codesight-scan.mjs" <path> once to seed .codesight/CODESIGHT.md.
c. Authored (you write, only if scan marked missing, existence-checked first):AGENTS.md — the single authoritative entry point: real build/test/run commands (from the detected stack + the local-dev branch), conventions, short architecture overview.CLAUDE.md — thin pointer to AGENTS.md (no duplicated content)..cursor/rules, GEMINI.md, .github/copilot-instructions.md, .clinerules, .opencode/…, .amazonq/…) — each a thin pointer to AGENTS.md.structure=mono: author <submodule>/CLAUDE.md per submodule (stack-specific)..codesight/CODESIGHT.md only if contextMode=static (otherwise the regen hook owns it).Re-score (CLI): run ... score <path> --profile <profile> and report the new % + remaining gaps. Note that functional scoring counts only WIRED hooks / FRESH indexes / host-matched CI — explain any item that stayed partial and why.
End with: 💡 /codepresso:aidlc-doctor <path> — re-check compliance anytime.
<Tool_Usage>
Bash for the aidlc-cli.mjs subcommands (always pass --profile after step 3) and for wiring hooks.AskUserQuestion for the 7-branch interview + preview confirmation.Read/Write for the profile JSON and authored files (Write only after existence check — non-destructive).
</Tool_Usage>npx claudepluginhub code-presso/codepresso-ai-plugin --plugin codepressoGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.