bootstrap-claude

File a new bug report in wiki/work/bugs/ with required-on-report fields and append it to the bug index

Audit all project functionality at a high level and produce a 2-3 minute demo run book plus a Marp slideshow

Orchestrator for parallel agent teams. When invoked with a roadmap file path, drives every item through the full tackle → uat-generate → uat-auto pipeline until complete. When invoked with a task file path, runs that pipeline once and stops.

Refresh codebase context via Serena memories

Append a new item (task link or inline) to an existing roadmap in wiki/work/roadmaps/, optionally under a named phase

Create a structured execution-plan roadmap in wiki/work/roadmaps/ via short Socratic Q&A — captures goal, phases, and the initial hybrid (task-link OR inline) checklist

Execute an outlined task file step-by-step with subagent delegation

Generate a dependency graph of active tasks showing which block others and which can run in parallel; also checks for unannotated implementations and auto-completes tasks

Trash a task and its related UAT files — moves them to archive/ and removes all active-index references

Autonomous-fix variant of /uat-auto — runs every test, diagnoses failures, applies fixes itself, and re-runs until green or attempts are exhausted. Intended for headless agents launched with --dangerously-skip-permissions.

Non-interactively run every test in a pending UAT file and auto-judge verdicts (headless, fail-closed)

Generate UAT tests in wiki/work/uat/ for a task

Update task, UAT, and project documentation files to reflect implementation work just completed

Analyze a feature in the current project and produce a SOLID-compliant standalone module extraction plan with interfaces for all dependencies; output is compatible with /port-feature

Turn markdown/research files into a self-contained interactive flashcard HTML page

Assess a feature or set of functionality in an external project and produce a concrete porting plan for adapting it into the current project

Comprehensive company research — mission, operations, leadership, products/services deep-dive, and ~5 years of recent news

Deep research on a topic using codebase analysis, library docs, and web search — writes the report plus its primary sources to raw/research/<slug>/

Interactively configure Serena language servers in .serena/project.yml

Analyse project–company fit — maps project implementation, style, and functionality against company mission, values, and likely needs; surfaces gaps and strengths

Close an in-progress bug — record root cause and resolution, require a regression test, then move it to archive/ (or delete it for late wontfix decisions)

Triage an open bug — set priority/assignee/tags/impact, then keep it in wiki/work/bugs/, advance it to in-progress, or reject it (wontfix/duplicate/cannot-reproduce) and archive it

Finalize a single proposed decision block; run E-C-A-D-R audit, supersession check, and flip status to accepted

Retire a requirement — set status to retired in frontmatter, document the reason, move to archive/, append a log entry.

Point at the first unchecked item(s) in a roadmap; create task files for inline placeholders; group items into parallelizable waves; auto-archive fully-checked roadmaps

Create a structured, execution-ready task file in wiki/work/tasks/

Skip UAT for a task — sets UAT and task status to skipped/done, removes index rows, auto-checkoffs roadmap

Walk through a pending UAT file test-by-test with the user

Cleanup tool — sweep for work items with terminal status that were not auto-archived by their originating skill, and move them to archive/

Health-check the wiki — find contradictions, stale claims, orphan pages, missing concept pages, missing cross-references, and never-ingested raw sources; fix only with approval

Rotate wiki/log.md into a timestamped archive file when it grows past ~500 lines; create a fresh log.md with an archive-pointer header

One-shot wiki cleanup — runs lint, archives terminal items across all families, and rotates the log if overgrown, in sequence with user confirmation at each phase

Assess the project's eval coverage against the 5-stage framework, identify gaps, and create new evals one at a time with user understanding and approval at each step

Generate a short, punchy elevator pitch for a project

Create a Decision Group file in wiki/work/decisions/ with one or more proposed decisions, table-only comparisons, and mermaid flowcharts

Audit the current app's functionality against expected functionality and produce a gap assessment as an approvable plan (runs in plan mode, delegates to subagents)

Compile a requirement into a running, production-grade, delightful system with minimal human time — by routing every claim to its cheapest re-runnable check and building the dominant oracle first. Use this whenever a project has a wiki/work/requirements/ REQ file to build or execute ("build this per the requirement", "execute the spec", "turn these requirements into a working system"), when scaffolding or resuming a greenfield build from a requirements document, or in a build's closing phase to harvest corrections back into the method. Pairs with frontend-design, run/verify, and claude-api.

Extract Architecturally Significant Requirements from an approved requirement, cross-check existing decisions, and propose Decision Group candidates for /decision-create

Audit an LLM/AI application for security vulnerabilities across 11 categories — internal posture (observability, rate limiting, access controls, HITL policy, benchmarking) and external threats (prompt injection, data leakage, output XSS, excessive agency, supply chain, token DoS). Run a full audit or a single named category.

Detect and remove AI writing patterns from any copy — blogs, social posts, emails, landing pages, ad copy, pitch decks, or internal documents — so the writing sounds like a real person thinking out loud instead of a language model producing persuasive text. Use this skill whenever someone needs to write or edit copy that should not read as AI-generated, whenever previous AI output has come back generic or hollow, or whenever a piece of writing feels technically correct but somehow off. Triggers include: "this sounds like AI," "make this sound more human," "edit this copy," "write a LinkedIn post," "write a blog," "rewrite this," "clean up this draft," or any request to produce or improve written marketing copy. Always apply the master prompt at the end of any copy prompt, never at the beginning.