giostriquer
- 34 skills
- 0 followers
- 6 hours ago last updated
- ▌ Push · giostriquerUse when committing and pushing changes in the project repository; detect the current branch, pull origin for that branch before committing, prefer the newest meaningful update in docs/change-log.md as the commit message source when available, otherwise derive the message from the primary diff, then stage all changes, commit, and push to the same current branch.
- ▌ Change Log · giostriquerUse when a change ships in a plugin version bump (workbench, toolkit) and needs its release-notes entry in docs/change-log.md. Not for repo-only work (structure, docs, and decisions, which live in docs/decisions/ and git history), formatting edits, or scratch work.
- ▌ Orchestrate · giostriquer bundleUse when the user hands over a set of work (a feature, a fix list, a refactor campaign) to complete autonomously by coordinating dispatched executors. This session decomposes, dispatches, judges, and iterates until the whole set is done. Not for single small tasks (dispatch those directly via codex-implement) or for work the user wants to drive step-by-step.
- ▌ Codex Implement · giostriquer bundleUse when dispatching implementation tasks to Codex CLI (GPT-5.5) as the coder while this session plans and judges, once a written plan or task brief exists and coding labor should be delegated to codex exec. Not for planning, review, or tasks that need conversation with the user.
- ▌ Writing Skills · giostriquer bundleUse when creating new skills, editing existing skills, auditing existing skills or verifying skills work before deployment
- ▌ Workbench Drift · giostriquer bundleUse when checking the workbench skill set (the workbench plugin's superpowers-derived process skills) against upstream (obra/superpowers), on demand ("check superpowers drift") or on a review cadence. Runs the bundled deterministic diff, then judges each upstream change against the manifest's recorded dispositions and recommends adopt / adapt / ignore, verdict-first. Never auto-applies upstream changes; advances the manifest's reviewed-commit pin only after the review completes; never commits.
- ▌ Audit · giostriquerUse for a requested investigation or audit; routine checks and known fixes stay in their ordinary workflow. Not for work that starts from an idea to build (that path grounds against the codebase and goes to brainstorming).
- ▌ Arch Map · giostriquer bundleUse when you need a visual architecture map of a codebase and no finished source document exists (an existing subsystem, a refactor in flight, or a proposed design) rendered as a self-contained HTML page. NOT for rendering a report that already exists, on disk or as findings reached in the conversation; that is html-report.
- ▌ Me Human · giostriquerAct as a human user dogfooding this system for real work (tries before asking, escalates on bugs, stops at the scope edge) and reports what got in the way. User-invoked only.
- ▌ Fix CI · giostriquerUse when the current branch's CI is failing ("CI is failing, take a look"), after a push to see the checks through to green, or when a red check needs diagnosing and fixing.
- ▌ Research · giostriquer bundleForward-looking research skill. Produces docs/research/-conforming notes across project-defined categories with consistent inputs, five-axis scoring, and a Horizon flag. Use when the user asks for a broad research pass on a category, wants to reassess existing seeds, or wants to surface net-new ideas grounded in a category's purpose anchor.
- ▌
- ▌ Doc Audit · giostriquer bundleRun a proactive audit of the project documentation surface. Surfaces gaps that diff-driven wiki-maintainer passes miss by construction, such as missing pages, broken links, undefined terms, missing decision records, orphaned files. Report-only; never edits docs directly. Use when the user asks for a doc audit, a coverage check, a drift sweep, or "what's missing in our docs."
- ▌ QA Sweep · giostriquerUse when running a broad QA or verification pass over a release, branch, feature, or app surface that splits into independent slices and deserves team-scale coverage against the real running artifact, on the user's ask (directly or as audit's team-sweep pick), never by default. NOT for a single code change (that is empirical-proof) or a single premise, ticket, or hunch (that is claim-check).
- ▌ HTML Report · giostriquerUse when turning a report, audit, review, or research/findings into a polished self-contained dark-themed HTML page (whether the source is a markdown document or material that exists only in the conversation) or when revising such a page through content tweaks, section changes, severity ordering, repo-style matching, or "this looks noisy / unreadable / ugly / off" feedback. For deriving an architectural representation from code, a diff, or a plan when no source document exists, use arch-map instead.
- ▌ Agent Audit · giostriquer bundleRun an advisory agent-governance audit or vigil consulting pass. Use when asked to audit project agents, review agent/skill/wrapper drift, check cross-host wrapper parity, review a proposed agent change, or decide whether a recurring workflow needs an agent, skill, or rule update.
- ▌ Self Audit · giostriquerUse when the session should look back at how the workbench process itself ran, rather than at the code it produced or the user's choices. User-invoked only.
- ▌ UI Demo Video · giostriquer bundleUse after UI work that can be verified visually, such as a new element, layout change, or user flow in a web app whose dev server a browser can drive. Not for API-only or non-visual changes, and not a test suite.
- ▌ Claim Check · giostriquerUse when you have a premise to verify before acting on it (a tracker ticket, a hunch you are carrying, or a bare question) and you want a deep, evidence-grounded investigation rather than a guess. Not for implementing the fix the premise calls for.
- ▌ Handoff Goal · giostriquerUse only for long-running work that must outlive this session, meaning a defined goal a fresh session pursues autonomously over many turns, or an existing goal contract that needs critique. Not for work this session can finish, and not for open-ended research.
- ▌ Handoff Review · giostriquer bundleUse when work on a branch needs a fresh, unbiased pair of eyes, whether a pre-PR review or a clean restart when this session's context has gone bad and a new session should independently verify what was already done and continue it. Produces a self-contained brief (task-vs-code, rules conformance, information leak, correctness) that re-derives the task from ticket + diff, never trusting the prior session. Default spawns a fresh reviewer (verify-only); `handoff`/`session` writes the brief to a scratch file for a new session (verify-only); `continue`/`resume` writes a verify-then-continue brief carrying current state, remaining work, and operating rules, pointing to handoff-goal for substantial forward work.
- ▌ Visual Advisor · giostriquer bundleUse when visual work needs art-direction judgment, AI-image prompt shaping, candidate critique, asset / animation pass selection, promotion decisions before runtime implementation, or when the current visual direction feels stuck on the same look and needs broader exploration.
- ▌ Get Pr Comments · giostriquerFetch and summarize review comments from the active pull request. Use when you need a concise, actionable summary of feedback on the active PR.
- ▌ Brainstorming · giostriquerUse when designing a feature or a refactor and when an idea carries questions the codebase can't answer (intent, preference, trade-offs). Not for confirmed small fixes or work whose design is already settled.
- ▌ Empirical Proof · giostriquerUse when a just-finished change touched a surface a real client can drive (an MCP tool, a REST endpoint, runnable app behavior, or a generator's emitted artifact) and the user asked for empirical verification, which includes a repo completion gate that requires driving the real artifact (offer it otherwise; never run it uninvited). NOT for a release- or branch-wide pass (qa-sweep), verifying a premise or ticket (claim-check), or driving an app to hunt for unknown bugs in a surface, which is ordinary session work rather than this protocol.
- ▌ Model Reference · giostriquerUse when a model choice needs grounding and is not already settled, or when the fleet changes. Not a step before every dispatch.
- ▌ Using Workbench · giostriquerUse before coding, auditing, planning, shipping, filing a PR, debugging or using any workbench skill.
- ▌ Adopt Global Rules · giostriquer bundleInstall the workshop's shipped global agent configuration onto this machine, additively; the payload is a per-host instruction document (CLAUDE.md / AGENTS.md), discrete rules, and output styles. Reconciles what is already there instead of overwriting it. User-invoked only.
- ▌ Epic Orchestration · giostriquerUse when running a multi-ticket epic whose implementation you delegate to other sessions the operator dispatches by hand, and when closing such an epic on a blind re-audit rather than an empty ticket list. This session owns the epic and never implements, commits, or opens the PR itself. Not for implementing a change yourself, and not for a single ticket.
- ▌ Code Quality Review · giostriquerUse for a strict or adversarial code quality review; required once a work-stream's implementation is complete, right before PR-or-merge, never mid-implementation.
- ▌ Systematic Debugging · giostriquer bundleUse for an unresolved failure requiring sustained investigation, such as intermittent behavior, unclear cross-component causes, or unsuccessful fixes, or when explicitly requested. Skip expected TDD RED, obvious localized fixes, and incidental out-of-scope bugs.
- ▌
- ▌ Test Driven Development · giostriquer bundleUse when implementing a feature or bugfix in a repo that has a test harness. Where the repo has no harness, skip without ceremony.
- ▌ Verification Before Completion · giostriquerUse when about to claim work is complete, fixed, or passing.