AI Agent Workflow Collection

Make a target project's own skills usable in the current session. Reads /.claude/skills/, symlinks each skill into ~/.claude/skills/ (so Claude Code hot-reloads them on the next turn), and records them in acquired-project-skills.json so the SessionEnd hook can remove them on exit. Invoke this at the START of working in any TARGETDIR that ships its own .claude/skills/ — including from the project-builder, develop, revise-brief, and define-use-case entry points — or run it standalone to pull in one project's skills. Skips (never clobbers) any name already present in ~/.claude/skills/ or any name that matches a workspace skill.

Check whether the current Claude session was launched with --dangerously-skip-permissions. Run ps aux, find the claude process, and report yes/no.

Capture deployment characteristics for both production and development — hosting, IaC, CI/CD, secrets, observability, local dev environment, containerization, hot reload, and seed data. Use when the project-builder agent is gathering deployment information.

Capture the monetization and distribution model of the project — business model, pricing tiers, target market, licensing, and any non-commercial framing. Use when the project-builder agent is gathering monetization information.

Capture the high-level definition of a project — problem, target users, value proposition, scope boundaries, non-goals, and success criteria. Use when the project-builder agent is gathering overview information.

Capture the quality and standards baseline — linting, formatting, testing strategy, security, accessibility, performance budgets, documentation, and observability. Use when the project-builder agent is gathering quality/standards information.

Capture the technology choices for the project — languages, frameworks, data stores, auth, and key libraries. Stack-neutral; presents options with tradeoffs. Use when the project-builder agent is gathering technology information.

Capture a single use case for a target project. Collects a free-form description from the user, produces a formalized version (summary, acceptance criteria, pitfalls), runs a clarifying-question loop, and saves the result as an incrementally-numbered Markdown file in the target project's use-cases/ folder. After saving, offers to define another use case or continue to implementation via the develop skill. Use when the user wants to document what the project should do, before or instead of jumping into implementation.

General principles for exploring, developing, editing, and refactoring code. Use for codebase analysis, implementation tasks, and code quality improvements.

Trigger and monitor GitHub Actions pipelines via the gh CLI — push-triggered workflowrun dispatch, tag-triggered release workflows, PR check rollups, and the polling patterns that reliably exit on terminal states. Includes the exact field names, expected success/failure outputs, and the foot-guns that silently leave polls spinning forever. Use whenever a Claude run needs to wait for CI / a release workflow to finish, or wants to read out a failed step's logs without dumping the full run.

Opt-in profile. AWS is the preferred cloud provider. Any AWS-based suggestion or solution MUST include a cost estimation table — per-service daily / monthly / yearly costs and cumulative totals that account for both newly proposed services AND pre-existing AWS services already in use by the project. Only invoke when PROJECTBRIEF.md's Profiles section lists this skill by name. Do NOT auto-invoke on arbitrary cloud or deployment discussions.

Opt-in profile. Provisions the java-class-call-scanning bytecode call-graph analyzer (https://github.com/HaroldHormaechea/java-class-call-scanning) for use by the dev-team agents on a Java project. Downloads the latest release jar to a per-user cache, registers the daemon as an MCP stdio server scoped to TARGETDIR, and documents per-role usage of the nine query operations (find-callers, find-callees, methods-in-class, methods-at-line, find-field-readers, find-field-writers, impact-of-diff, tests-for-diff, refresh-index). Only invoke when PROJECTBRIEF.md's Profiles section lists this skill by name. Do NOT auto-invoke on arbitrary Java work.

Opt-in profile. Opinionated Java database-access conventions — DTO projections over entity queries, bulk retrieval over iterative round-trips, parameterized queries only (never string concatenation), and Hibernate as the JPA implementation. Only invoke when PROJECTBRIEF.md's Profiles section lists this skill by name. Do NOT auto-invoke on arbitrary Java or database-related work.

Opt-in profile. Opinionated Java server-side architecture conventions — Gradle on its latest stable version, latest stable Spring Boot, latest Java LTS, repositories return DTOs (never JPA entities) to upper layers, internal DTOs stay internal (API layer has its own request/response DTOs), and a strict Controller/Job → Facade → Service → Repository call chain with transactions owned by the facade. Only invoke when PROJECTBRIEF.md's Profiles section lists this skill by name. Do NOT auto-invoke on arbitrary Java or Spring work.

Update one or more sections of an existing PROJECTBRIEF.md in a target project folder without re-scaffolding. Lets the user pick which section(s) to refresh (overview, monetization, technologies, architecture, quality-and-profiles, deployment) and delegates to the project-builder subagent to re-run only those define- skills. Keeps the YAML frontmatter in sync and preserves untouched sections. Use when the target project already has a brief and the user wants to evolve it; use project-builder for initial scaffolding or develop to build features.

Safely wait for and monitor any external-consequence outcome — CI runs, deploys, queue drains, container/service health, files appearing on disk — using the general "pre-flight the polled command, cap the loop, check exit code every iteration" pattern that prevents silent infinite stalls. Documents the reusable waituntil wrapper plus a concrete GitHub Actions instantiation (push / tag / PR-triggered runs, terminal-state field vocabularies, log-failed extraction) with every foot-gun that silently breaks polls. Use whenever a Claude run needs to wait for an asynchronous outcome before its next step.