Last updated: 2026-09-23
Auto-generated — run /skills-dashboard in Claude Code to refresh. Contents reflect files that exist on disk at time of last run; no manual inventory needed.
Source type: slash commands (~/.claude/commands/) except task-from-chat (user skill).
| Command | How to invoke | Purpose |
|---|
| task-start | /task-start [file] | First session on a task — read, execute, append first Claude Response |
| task-continue | /task-continue [file] | Process latest Talbot Response, append new Claude Response |
| task-complete | /task-complete [file] | Close task: extract lessons, log/delete, commit automatically |
| task-compact | /task-compact [file] | Shrink a long task file (≥40 KB) to a summary + verbatim last round |
| task-prep | /task-prep [file] | Hydrate task with full context; clarity check; sets status: ready |
| task-from-chat | ”convert this to a task”, /task-from-chat | Convert chat-discovered task into SMTM task file |
| project-start | /project-start [name] | Initialize new KB project structure |
| project-continue | /project-continue [name] | Continue any project. Finite: processes the Talbot Response in NEXT-STEPS.md. Ongoing: opens the active task named in ROADMAP ## Now (v5.9) |
| project-task-complete | /project-task-complete | End project session: write log, tick STATUS.md, update _tmp.md |
| triage-inbox | /triage-inbox [clippings] [--batch N] | Triage Clippings into Capture-Routing: rate 0–10, TLDR, propose route, batch approve (human-gated) |
| Skill | How to invoke | Purpose |
|---|
brainstorming [Plugin] | ”let’s brainstorm”, “explore options” | Explore intent and design before implementing |
writing-plans [Plugin] | ”write a plan”, “plan this” | Create implementation plan from spec/requirements |
executing-plans [Plugin] | ”execute the plan” | Run a written plan with review checkpoints |
subagent-driven-development [Plugin] | auto-triggered | Parallelize independent tasks via subagents |
dispatching-parallel-agents [Plugin] | 2+ independent tasks | Dispatch independent tasks to parallel agents |
using-git-worktrees [Plugin] | starting isolated feature work | Ensure an isolated workspace via git worktree |
test-driven-development [Plugin] | auto-triggered on features/bugfixes | TDD workflow before writing implementation code |
create-review-guide [Global] | /create-review-guide <dev-project-path> | Generate/refresh Review-<Project>.md in the repo — run commands, check punch list, static proof, gaps, tests; zero-friction clickable links |
| Skill | How to invoke | Purpose |
|---|
systematic-debugging [Plugin] | auto-triggered on bugs/failures | Debug before proposing fixes |
verification-before-completion [Plugin] | auto-triggered before “done” | Require evidence before claiming success |
requesting-code-review [Plugin] | auto-triggered after implementation | Validate work meets requirements |
receiving-code-review [Plugin] | auto-triggered on review feedback | Handle review feedback with rigor |
finishing-a-development-branch [Plugin] | auto-triggered when work is done | Decide merge/PR/cleanup strategy |
| Skill | How to invoke | Purpose |
|---|
using-superpowers [Plugin] | auto-triggered at session start | Establish skill discovery and usage rules |
| session-save | ”save context”, “checkpoint”, /session-save | Emergency save — write recovery file for next session |
| session-continue | ”resume session”, “load context”, /session-continue | Load a session-save file and resume autonomously |
| Skill | How to invoke | Purpose |
|---|
| research | /research [path] | Research a topic from a task file; appends cited report |
| graphify | /graphify [input] | Convert any input (code, docs, papers, images, video) into a knowledge graph; query an existing graphify-out/ |
| hyperresearch | /hyperresearch <query> | Deep multi-agent research pipeline (tier-adaptive: light/full/dissertation) into its own vault, rooted at ~/hyperresearch/ (moved from ~ 2026-09-23). Start /hyperresearch only from a session whose working directory is ~/hyperresearch: the hyperresearch-root-guard hook blocks vault creation anywhere else (KB-OS-cleanup); heavier and slower than /research, reserve for high-stakes questions worth the session-usage cost. Piloted 2026-09-15/17, Core/AI/Tasks/ai-config-deep-research-tools.md |
| Skill | How to invoke | Purpose |
|---|
| financial-dashboard | /financial-dashboard [asset class + date range] | Generate standalone dark-themed HTML dashboard: historical rolling returns (nominal vs. real) for any asset class, CA + US regions, dual 10/30-yr horizons |
Reference example: ~/.claude/skills/financial-dashboard/references/precious_metals_investing_dashboard.html
| Skill | How to invoke | Purpose |
|---|
design-context [Global] | /design-context <web|app|print|deck|report> — load before any design work | Talbot’s brand + F.A.S.T. constraints per target. Vault SSOT Core/Processes/Design/, bundled into references/ at deploy time. F.A.S.T. outranks conventional design convention; hands off to the craft skills below |
frontend-design [Plugin] | auto-triggered on web UI work | Create distinctive, production-grade frontend interfaces |
artifact-design [Plugin] | auto-triggered before writing an Artifact | Design fundamentals for published Artifact pages |
dataviz [Plugin] | auto-triggered before any chart | Charts, dashboards, palettes, stat tiles |
| Skill | How to invoke | Purpose |
|---|
| skills-dashboard | /skills-dashboard | Regenerate this dashboard from disk scan |
| skill-security-review | /skill-security-review, “check this skill for security” | Pre-use security gate for a new/external skill — prompt injection, code exec, supply chain; stamps security-tested date |
skill-creator [Plugin] | ”create a skill”, “improve this skill” | Create/iterate on skills with evals and description optimization |
writing-skills [Plugin] | auto-triggered when creating/editing skills | Ensure skill quality before deployment |
| Skill | How to invoke | Purpose |
|---|
caveman [Plugin] | /caveman, “caveman mode” | Compress AI output ~65-75%; ultra-concise responses |
cavecrew [Plugin] | ”delegate to subagent”, “use cavecrew” | Decision guide for delegating to caveman-style subagents |
caveman-commit [Plugin] | /caveman-commit | Write git commit messages in caveman style |
caveman-review [Plugin] | /caveman-review | Code review in compressed caveman output |
caveman-compress [Plugin] | /caveman-compress FILEPATH | Compress natural-language memory files into caveman format |
caveman-help [Plugin] | /caveman-help | Quick-reference card for all caveman modes and commands |
| caveman-stats | /caveman-stats | Show token savings stats for session |
Auto-activates via SessionStart hook. Modes: lite, full (default), ultra, wenyan.
caveman-stats has a user command override at ~/.claude/commands/caveman-stats.md (not in ai-config — plugin-owned, see plugins-manifest.md).
| Command | Invoke | Purpose |
|---|
| content-expand | /content-expand [vault-path] | Phase 0.5 content pass on an Obsidian vault — tighten headlines, sharpen copy, verify CTA alignment |
| cleanup-markdown | /cleanup-markdown | Fix markdown that renders wrong in Obsidian — $-sign code blocks (AI-chat exports) + indented-outline code blocks (Ecco/Dynalist pastes) |
| filter-focus-refresh | /filter-focus-refresh | Refresh Filter+Focus.md — extract new ideas, rank against existing rocks, update on approval |
Dept: Accounting (A1 — drafts, CEO approves). Neither tool posts to QBO; both stop at a reviewable artifact. Code: ~/utils/acct-qbo/. Vendor SSOT: Core/Accounting/Config/vendors/.
| Command | Invoke | Purpose |
|---|
| hst-assistance | /hst-assistance --from <YYYY-MM> --to <YYYY-MM> | The routine one. Reads BMO CC + chequing statements → one QBO manual-entry artifact in work order (expenses by vendor, CC by month + Pay CC Bill, issues) |
| process-expenses | /process-expenses <pdf> | Occasional. Reads a vendor bill → structured fields + arithmetic reconciliation check. Use when a recurring bill’s rate changes (record the new split in rogers.yaml) or a new bill vendor appears |
The two read different inputs and neither calls the other: /hst-assistance reads Statements/, /process-expenses reads Bills/.
Supporting util: pdftext (~/utils/pdftext) — deterministic PDF text extraction for figure work; see pdftext. Not MarkItDown, per AGENTS.md “Document extraction — pick by job”.
Cadence-tagged health-check skills. SSOT for cadence + last-run tracking: Core/Processes/Projects/KB-OS/Routines.md. session-start-inject.sh surfaces overdue rows automatically.
| Skill | How to invoke | Purpose |
|---|
| audit-kb-os | /audit-kb-os | Monthly — vault health: broken wikilinks, ambiguous same-filename links, dept Tasks/Logs hygiene, folder-note gaps |
| audit-skills | /audit-skills | Monthly — skill library health: security check done, stale hardcoded paths, folder-assumption drift, .obsidian/ config drift, Skills Dashboard sync |
| audit-blocked | /audit-blocked | Fortnightly — surface every open item waiting on Talbot, gates first, aged |
| audit-backups | /audit-backups | Monthly — verify my_backup/job_monitor jobs actually ran and succeeded, not just configured |
| audit-inbox | /audit-inbox | Monthly — flag dept Inbox.md files with an aging or growing un-triaged backlog |
| Skill | How to invoke | Purpose |
|---|
| dept-init | /dept-init <Biz>/<Dept> | Initialize a dept for active use — creates Tasks/, Inbox.md, Logs/ |
| scaffold-util | /scaffold-util [name] [desc] | Scaffold a new FSS utility (WSL or Windows) with full project structure |
| setup-ongoing-project | /setup-ongoing-project {name-or-path} | Scaffold or update an evergreen “ongoing project” with standard AI context files — never overwrites existing files |
Project-scoped commands live in .claude/commands/ within the project repo — not auto-discovered by global scan.
| Command | Invoke | Purpose |
|---|
| test-all | /test-all | Full test suite: unit + TS + lint + 7 Playwright projects |
Deployment: Use the web-deploy utility (~/utils/web/web-deploy/) directly — no slash command.
Generated by /skills-dashboard · Source: ~/.claude/skills/skills-dashboard/SKILL.md