tapestry-patterns

Plugin tapestry-patterns from marketplace tapestry in Lizo-RoadTown/tapestry. Commands read from the repo's manifest — 3 days ago.

Skills (17)

• agentic-skill-design

  Meta-skill for designing skills that DECIDE and EXECUTE rather than ask the user permission for every choice. Use when authoring a new skill or rewriting an existing skill that asks too many questions. Captures the PROBE → DECIDE → ACT → REPORT pattern Liz wants for stack-with-conditions requests.

• changelog-entry

  Use at the end of substantive work — a migration, a new capability/agent/skill/endpoint, a fix worth remembering, a scope decision (absorbed/deferred/retired), or a version bump consumers act on — to write a `docs/changelog/` entry so the change is findable later without reconstructing it from git.

• concrete-rule

  Use when designing or fixing a system invariant that, if violated, causes a core failure of the whole project. Forces failure-mode anticipation + defense-in-depth + loud visibility into every concrete rule, so a single config drift, a forgotten plugin update, or a service drop cannot silently break

• deep-research-pattern

  Architectural pattern for multi-agent deep research with strict context isolation. Use when designing or extending the research subagents under subagents/ — covers shallow vs. deep mode, the three-role topology, and context boundaries.

• design-evaluation

  Evaluate a design question with multiple options across the dimensions that matter for the project — produces a tradeoff matrix, scores each option, and recommends a path (often a hybrid). Use when the user asks "which approach", "should we use X or Y", "compare these styles", or shares competitor s

• device-coordination

  Use when about to make a change that differs per machine or crosses the operator's devices — bumping/updating a Claude Code plugin, running a machine catch-up, editing installed_plugins.json / ~/.claude.json / ~/.claude/settings.json / a project .mcp.json, or any per-repo config — on a repo shared a

• document-parsing

  Convert PDFs, Word docs, PowerPoint, Excel, and scanned images into LLM-friendly markdown. Use when the agent needs to read source documents — research papers, reports, slide decks, financial filings — that aren't already in plain text.

• documentation

  Plan, write, and audit project documentation using the Diátaxis framework, Architecture Decision Records (ADRs), and the docs-as-code workflow. Use whenever a task involves creating, reviewing, restructuring, or auditing documentation — including READMEs, API references, tutorials, design docs, or i

• id-rename-blast-radius

  Before renaming an identifier that is simultaneously a code key, a filesystem path, and a persisted namespace, enumerate everything derived from it. Use when changing a registry key, tenant id, course/project slug, storage prefix, folder name, or route segment - especially when the "rename" looks li

• layered-explanation

  Use BEFORE every explanation of infrastructure, architecture, tooling, library choice, telemetry stacks, CI pipelines, or any technical concept where the user needs to make a decision. Structures the response as four progressively-deeper layers (ELI5 metaphor → quick-reference table or diagram → dep

• open-source-documentation

  Maintain documentation for an open-source project so contributors can onboard, design decisions are auditable, and per-section docs stay current. Defines the docs/ tree (concepts / how-to / reference / decisions / proposals), the ADR pattern for design decisions, when to update what, and the dual-mo

• periodic-architectural-checkin

  Step back from active work and audit drift between current project state and the original goal. Surfaces (1) work that's happened, (2) where the work has drifted from the goal, (3) categorization / ownership gaps where no agent owns a domain, and (4) what to discuss with the operator. Use after N su

• plugin-authoring

  Use when a project's shape calls for its own Claude Code plugin — scaffold a personalized plugin (project-specific guards + the user's OWN backend endpoints), test it, and publish it to the user's OWN marketplace. The universal tapestry-discipline plugin is the always-on baseline; this builds the pr

• project-launch

  Use when creating ANY new project or repo for the operator, in any wording — "launch a project", "start a repo", "new project called X", "scaffold me a...", "set up a project for..." — or when a session discovers an existing repo that was never onboarded (missing .project-intelligence/, flagged by t

• proposal-authoring

  Author a design proposal in Make_Skills's house style — fixed section layout, project tone (state what is, not what it isn't), two-mode notes, open questions, references. Use when starting a new file under docs/proposals/, when transcribing a research synthesis into a proposal, or when reviewing a d

• seed-leftover-audit

  Sweep a freshly scaffolded or forked repo for strings inherited from the template or the previous instance - placeholder tokens, the old project's slug, last term's names, stale memory-boundary claims. Use immediately after `tapestry init`, after cloning a `*-starter` seed, after forking a working r

• template-inheritance-check

  After forking a WORKING instance (not a blank template) to start a new one, find the domain assumptions the source's data made that the new instance's data breaks. Use when a repo was copied from last term's app, the last client's site, or a sibling service, and the new instance's shape differs - di

Subagents (9)

• agentic-upskilling

  Documentation reference for the deployed self-observer cron service. NOT an interactive subagent — the actual implementation runs as a Render cron (every 6h) at the-loom/services/self-observer/, scanning platform-owned repos for category-drift candidates and emitting them to architecture-registry. R

• drift-watcher

  Use when starting a long execution (>20 min, multi-repo, bound by binding rules) where the primary agent has a history of drifting — per-repo solutions when cross-repo required, forgetting binding framing, skipping PROBE-before-asserting, re-creating duplicates just deleted, inventing things when me

• eval-deep-research

  Use when scoring a deepagents-produced research report set against the deep_research_bench (DRB) — 100 PhD-level tasks scored by RACE (report quality) + FACT (citation grounding) judges. Clones harness on demand, runs both scorers, parses outputs, emits structured score summary.

• infrastructure-mapping

  Use when auditing a codebase for the first time, after acquiring a repo, before a non-trivial refactor, or when the operator says "the system feels unstructured." Produces a module table (cohesion + instability), an interface table with signal-felt-by-Claude-vs-by-user as the load-bearing column, a

• lessons-learned

  Use when the operator wants the system to "get sharper" or after a long working session. Walks prior chat transcripts to find systematic friction patterns (misunderstandings, recurring info needs, operator corrections), then crystallizes them into intake forms + memory updates so future invocations

• next-actions-planning

  Use when the operator asks "what's next?" or after a major piece of work lands. PROBEs the repo + git + open proposals + recent memory + last 20 operator messages; SCOREs candidates on Blocking + User-pull + Effort; WRITEs a plan file at docs/plans/YYYY-MM-DD-next-actions.md; RETURNs a 3-bullet top

• orchestration-cataloging

  Use when the operator asks "what should I make reusable" / "what patterns am I repeating" or after several similar tasks ship in a row. Identifies recurring work patterns in recent build (research bursts, proposal writing, schema migrations, etc.) and recommends turning the high-frequency ones into

• roadmap-maintenance

  Use when work ships, gets blocked, or a new capability becomes worth tracking, and ROADMAP.md needs to reflect it. Reads ROADMAP.md, verifies the claimed evidence (commit/PR/file/test), then makes ONE minimal, evidence-backed edit — flip a row's status or append a new row. Never updates speculativel

• web-app-scaffold

  Use when the operator wants a chat UI / dashboard / frontend "on my website" / "live" / "hooked to my domain". PROBES existing frontend + API + authenticated tools; DECIDES the stack with defensible defaults (Vercel + Next.js 14 + Tailwind unless probe disconfirms); ACTS (scaffold + install + smoke

Similar plugins