claude-tweaks-help

name	claude-tweaks:help
description	Use when you need a quick reference for available commands, want to see workflow status, or need a recommendation for what to do next.

Interaction style: Present single decisions via the AskUserQuestion tool (options with one marked Recommended) instead of a plain-text numbered list. For multi-item decisions, render a batch table with recommended actions pre-filled, then capture the apply-all/override decision via one AskUserQuestion call. Never make more than one AskUserQuestion call per logical decision — resolve each before showing the next. End skills with a ## Next Actions block rendered via AskUserQuestion (context-specific options, one recommended), not a navigation menu.

Help — Quick Reference + Workflow Dashboard

One-stop reference and status dashboard for the workflow system. Combines command help, pipeline scanning, and next-step recommendations.

/claude-tweaks:capture → /claude-tweaks:challenge → /superpowers:brainstorming → /claude-tweaks:specify → /claude-tweaks:build → /claude-tweaks:review → /claude-tweaks:wrap-up
    ↑                                                                                                                                            |
    └──────────────────────────────────── [ /claude-tweaks:help ] (dashboard + reference) ←──────────────────────────────────────────────────────┘
                                            ^^^^ YOU ARE HERE ^^^^

When to Use

• You need a quick reference for available commands and their syntax
• You want to see what's in the pipeline and what needs attention
• You're unsure what to do next in the workflow
• You want a recommendation based on current pipeline state
• After completing a batch of work, to check what's unblocked

Input

$ARGUMENTS controls what to show:

Argument	Behavior
(none)	Cheat sheet + status scan + recommendation
status	Pipeline status scan only
commands	Quick reference cheat sheet only — read reference-card.md in this skill's directory and present its contents
spec number or topic	Targeted status for that specific spec/topic

Default (no arg) is the cheat-sheet-first dashboard. The status scan (Section 2) is moderately expensive — when the user just wants command syntax, the commands arg avoids it.

---

Section 1: Quick Reference Cheat Sheet

(Skip if $ARGUMENTS = status)

For the canonical cheat sheet — lifecycle, component, and utility commands; common workflows; artifact lifecycle; bookend architecture summary — read reference-card.md in this skill's directory. Rendering rule: present the Lifecycle / Component / Utility command tables verbatim. Collapse the Common Workflows section to a list of workflow headings (e.g., "New repo bootstrap", "Full pipeline", "Resume from step") unless the user explicitly asked for "examples" or "workflows" in $ARGUMENTS — in that case, render the workflows verbatim too. Always keep the Artifact Lifecycle and Bookend Architecture summary verbatim. The reference card is the single source of truth for the command catalog; do not maintain a separate copy here.

---

Section 2: Pipeline Status Scan

(Skip if $ARGUMENTS = commands)

Read status-scan.md in this skill's directory for the full parallel-dispatch procedure (Stages 1-7 including sub-stages 4.5 and 4.6; dispatch contract, agent template, and dashboard rendering). The orchestrator dispatches all seven stages in parallel and assembles the dashboard after all agents complete.

---

Section 3: Recommendation

(Always included unless $ARGUMENTS = commands)

Priority Order

1. Current PR blocked — the current branch's open PR has failing CI, CHANGES_REQUESTED, or unresolved review threads (Stage 4.5). PR feedback is the most perishable work in the system — reviewer context decays fastest and it blocks in-flight work from merging. Recommend fixing CI, addressing threads, or resuming /claude-tweaks:build before anything below.
2. Specs awaiting review — review completed work before it goes stale
3. Specs awaiting wrap-up — wrap up reviewed work (captures learnings while fresh)
4. Records building (bot:in-progress, Stage 1) — finish what's started before starting new work
5. Design docs unspecified — specify before building (don't let designs go stale)
6. Parked records with met triggers — promote via /claude-tweaks:specify before starting new work
7. Records pending authorization (ready, not yet granted, Stage 1) — recommend /claude-tweaks:triage to review and grant auto:build
8. Authorized records (Stage 1) — recommend /claude-tweaks:dispatch (headless) or /claude-tweaks:build #{n} (direct) for the highest-priority one with met prerequisites
9. Backlog review — if the backlog is stale or has 10+ records, suggest /claude-tweaks:tidy before new brainstorming
10. Challenge + Brainstorming — if the pipeline is empty, suggest brainstorming from a backlog record; if its title has baked-in assumptions, run /claude-tweaks:challenge first, then /superpowers:brainstorming
11. Nothing to do — if everything is clean, say so

Tie-Breaking

When multiple records are ready to build (no priority:* label to sort by, or several share the same priority):

• Prefer records that unblock other records (check the dependency graph — Blocked by #N / blockedBy)
• Prefer records with no file overlap with building work (avoids conflicts — see status-scan.md Stage 1's conflict detection)
• Prefer smaller records (effort:low, faster feedback loop)
• Prefer records with existing plans (less setup needed)

Detecting Items That Need /claude-tweaks:challenge

A backlog record likely needs debiasing when its title or body:

• Names a specific technology as the solution (e.g., "Add Redis caching" instead of "Improve response times")
• Frames the problem as a solution (e.g., "Build a microservice for X" vs "X is too slow")
• Contains strong assumptions about the approach without exploring alternatives

Mode recommendation: Records with mild assumption signals (slightly solution-oriented phrasing, but the problem space is mostly clear) → recommend /claude-tweaks:challenge quick {topic}. Records with strong solution-baking or multiple competing assumptions → recommend full /claude-tweaks:challenge {topic}.

Present Recommendation

Render the recommendation as the ## Next Actions block below via one AskUserQuestion call — {recommended command} becomes option 1, labeled with a short name suffixed (Recommended), with 1-3 alternatives drawn from the priority order.

Next Actions

Call AskUserQuestion:

• question: "What's next?", header: "Next step", multiSelect: false
• Option 1 — label: short name of the recommended command suffixed (Recommended), description: the full command with parameters + rationale
• Option 2 (and optional option 3) — same shape, for each alternative drawn from the priority order

Component-Skill Contract

/claude-tweaks:help is a standalone-only dashboard skill — it is not invoked by any parent skill in the workflow. There is no PIPELINE_RUN_DIR signal, no parent context to defer to, and the ## Next Actions block always renders. If a future parent skill ever invokes /help (e.g., a "show me the pipeline" dispatch), the parent must update this contract before that lands; until then, treat parent invocation as not applicable.

Anti-Patterns

Pattern	Why It Fails
Running a full scan when user just needs command syntax	Wastes time — respect the commands argument
Recommending new work when specs await review	Finish in-progress work first — stale reviews lose context
Recommending new work while the current PR has unresolved feedback or failing checks	In-flight work rots fastest — reviewer context decays and merge conflicts accumulate. The pipeline picture is incomplete without PR state.
Skipping the backlog scan	Stale backlog records create noise and slow down the pipeline
Not checking for baked-in assumptions	Solution-oriented backlog records bypass the debiasing step
Triaging backlog records from /help instead of handing off to /tidy	/help is a read-only dashboard — it reports status and recommends next steps. If the user wants to delete, promote, absorb, or defer backlog records, hand off to /claude-tweaks:tidy. Do not improvise an ad-hoc walkthrough.

For a detailed explanation of how context flows between skills via artifacts, read context-flow.md in this skill's directory.

Relationship to Other Skills

Skill	Relationship
/claude-tweaks:capture	Feeds items that /claude-tweaks:help surfaces
/claude-tweaks:challenge	/claude-tweaks:help flags items with baked-in assumptions for debiasing
/claude-tweaks:specify	/claude-tweaks:help flags unspecified design docs and uses Key Files from open records for implicit dependency detection (Stage 1's conflict detection)
/claude-tweaks:build	/claude-tweaks:help recommends which spec to build
/claude-tweaks:test	/claude-tweaks:help can recommend /test when code changes exist but no review is warranted
/claude-tweaks:review	/claude-tweaks:help flags specs awaiting review
/claude-tweaks:wrap-up	/claude-tweaks:help flags specs awaiting wrap-up
/claude-tweaks:tidy	/claude-tweaks:help suggests /claude-tweaks:tidy when maintenance is needed
/claude-tweaks:triage	Surfaces pending-authorization count, bot:blocked count, and rolling auto-merge count on the dashboard (Stage 4.6, triage-queue scope) — the reciprocal of triage/SKILL.md's own /claude-tweaks:help row.
/claude-tweaks:review-backlog	Reciprocal: could surface an "N unscored records" nudge on the dashboard — the reciprocal of this skill already surfacing /claude-tweaks:triage's pending-authorization count.
/claude-tweaks:demo	Surfaces the demo:pending count on the dashboard (Stage 4.7, acceptance-queue scope) — the reciprocal of demo/SKILL.md's own /claude-tweaks:help row.
/claude-tweaks:dispatch	/claude-tweaks:help surfaces the authorized and building counts dispatch acts on (Stage 1) — the reciprocal of dispatch/SKILL.md's own /claude-tweaks:help row.
/claude-tweaks:flow	/claude-tweaks:help lists /claude-tweaks:flow as an automation option for ready specs
/claude-tweaks:browse	Utility skill — /claude-tweaks:help lists it in the utility skills table
/claude-tweaks:stories	Lifecycle skill — /claude-tweaks:help lists it between /test and /review
/claude-tweaks:ledger	Utility skill — /claude-tweaks:help lists it in the utility skills table and scans for active ledgers with open items
/claude-tweaks:reflect	Component skill — /claude-tweaks:help lists it in the component skills table
/claude-tweaks:simplify	Component skill — /claude-tweaks:help lists it in the component skills table
/claude-tweaks:deepen	Component skill — /claude-tweaks:help lists it in the component skills table
/claude-tweaks:journeys	Component skill — /claude-tweaks:help lists it in the component skills table
/claude-tweaks:visual-review	Component skill — /claude-tweaks:help lists it in the component skills table
/claude-tweaks:visualize	Component skill — /claude-tweaks:help lists it in the component skills table
/claude-tweaks:init	/init configures the workflow system that /help reports on — /help reads the same artifact paths /init bootstraps (the work-record queue, plans, docs registry)
/claude-tweaks:version	/version prints the installed plugin version; /help surfaces version-aware command syntax and points at /version for the canonical answer
/claude-tweaks:design	Utility wrapper — /help lists it in the utility skills table. /design is invoked by /build (Common Step 1.7 pre-build), /test (Step 1.5 CLI gate), /review (Step 6.5 advisory pass), /flow (polish phase), and /visual-review; standalone usage is rare.
/claude-tweaks:research	Utility skill — /help lists it in the utility skills table. /research has no fixed lifecycle position; /help may surface it as an option when a backlog record or pending spec would benefit from prior-art research.
_shared/auto-mode-contract.md	Single source of truth for auto-mode behavior — read before adding any auto-mode handling to /help (e.g., if a future status scan ever auto-resolves recommendations)
_shared/github-pr-scan.md	Stage 4.5 scans the current branch's PR per this shared procedure (current-pr scope); Stage 4.6 scans the triage queue (triage-queue scope); Stage 4.7 scans the acceptance queue (acceptance-queue scope) — detection ladder, exact gh/GraphQL commands, output contract, severity mapping
_shared/work-record.md	Taxonomy home — the seven-axis label contract (stage / grants / bot state / origin / scoring / type / acceptance) Stage 1's record scan reads. /help has no row in the permission matrix — it never adds or removes a label, only queries the taxonomy every other row writes to.