xxthunder
UserMy reusable agentic skills marketplace
Categories
Indexed Skills (9)
architecture-scan
Derive a repository's real structure from its code and compare it against the architecture document. Produces a bootstrap proposal when no document exists, and a drift report when one does. Read-only — it proposes, and design-record applies. Trigger with: 'check the docs against reality', 'is the architecture doc still true', 'bootstrap the architecture doc', 'has the architecture drifted'.
backlog-ops
Apply lifecycle operations to a backlog item — pull (Open → In Progress), tick an acceptance or UAT criterion, or close an item (→ Done) with automatic epic-status cascade. Stages file edits only; never produces its own commit. Trigger with: 'start XAS-025', 'pull XAS-025', 'mark AC 2 done on XAS-025', 'tick UAT 1 on XAS-025', 'close XAS-025', 'complete XAS-025'.
design-record
Record a design decision as an ADR, or update the living architecture document, in a repo that keeps a record of how it is built and why. Drafts the ADR body from the conversation, allocates the next number, maintains the ADR index, and edits named sections of architecture.md. Stages file edits only; never produces its own commit. Trigger with: 'record an ADR', 'document this decision', 'we decided X — write it up', 'why did we choose Y — capture it', 'update the architecture doc'.
refinement
Start a backlog refinement session to discuss the project mission, review backlog items, prioritize work, and align on next steps. Also bootstraps the backlog structure in greenfield projects. Trigger with: 'another refinement session', 'let's refine', 'refinement time', 'backlog refinement', or similar requests to discuss project direction and priorities.
commit-helper
Create a conventional commit after running pre-commit checks (unit tests, plus integration tests when source dependencies change, plus linting). Use when the user asks to commit, says 'let's commit', 'commit this', 'create a commit', stages changes for commit, or finishes a unit of work that should land. Enforces conventional-commit format, atomic commits, and AI co-author attribution. Do not trigger on incidental mentions of 'commit' in unrelated conversation.
retrospective
Incident-driven learning and guideline improvement. Use when the user expresses dissatisfaction with Claude's recent work, approach, or decisions. Triggers include: 'I am not happy', 'this is not what I wanted', 'you shouldn't have done that', 'that's wrong', 'this is incorrect', or a 'why did you X?' question delivered with a clearly critical / corrective tone. Do NOT trigger on neutral 'why did you' clarifying questions where the user is simply curious about a decision. Captures what went wrong and updates project guidelines to prevent recurrence.
naps2-scan
Entry point for scanning paperwork with NAPS2. Drives a document scan end-to-end: invoke NAPS2.Console.exe to scan with OCR, chain into simplex-merge when scanning a double-sided document on a simplex scanner, and propose a content-derived filename. Use for ANY request to scan paperwork or digitize a document, including conversational variants like "scan this", "let's scan", "scan another", "scan again", "scan the next one", "new scan", "start scanning", "digitize this receipt/letter/invoice/bill/statement", or "run NAPS2". Also handles batch mode — scanning a whole stack of separate documents in one pass ("scan this stack", "batch scan", "these are multiple documents") — by chaining into `split-batch` after the scan to separate the bundle into one PDF per document. Also triggers when the user already has fresh scan output and wants to continue the pipeline (OCR, merge, name, rename). This is the starting-point skill for scanning; `simplex-merge` is post-processing only and should not be picked for new-scan r
simplex-merge
Merge two existing PDF files (odd pages + even pages) into one correctly ordered document. Post-processing only — this skill does NOT drive a scanner. Use ONLY when the user already has two PDF files in the working directory that represent the fronts and backs of a double-sided document. Typical cues: filenames containing "ungerade"/"gerade" (German), "odd"/"even", or "front"/"back", or the user explicitly asks to merge, interleave, or combine two existing PDFs. Do NOT use this skill to start a new scan or to digitize paperwork — for scanning, use the `naps2-scan` skill instead.
split-batch
Split a single multi-document PDF into one PDF per detected document. Post-processing for stacks of paperwork scanned in one pass: detect document boundaries from page content (page-number resets, letterhead/sender changes, address blocks) or from blank separator sheets, present a proposed split map, and emit one PDF per document after the user confirms. Use when the user has a searchable PDF that bundles several documents and says things like "split this batch", "split this stack", "these are multiple documents", "I scanned a pile", "separate these documents", or when `naps2-scan` hands off a batch-mode scan. Requires a searchable (OCR'd) PDF as input — it does NOT scan. For scanning, use `naps2-scan`.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.