Claude Skills by tomzx

Run the full software development lifecycle pipeline, from issue creation through implementation, documentation, and learnings capture.

Search a GitHub repository for potentially already reported issues matching a free-text problem description. Use when the user describes a bug, error, or feature they want to report, before creating a new issue. Triggers on "is this already reported", "search existing issues", "check if someone reported", or when the user describes a problem and wants to know if an issue exists.

Select the next issue to work on from a GitHub project's priority view, constrained to work also planned in the roadmap view. Use when the user says /select-issue, "what should I work on", "pick the next issue", or wants the next planned high-priority issue.

End-of-session checklist covering code quality (tests, docs, specs, simplicity, observability) and higher-level concerns (architectural fit, breaking changes, rollback, technical debt, communication).

Turn AI agent sessions archived by agentsview into durable memory. Reads the local agentsview SQLite archive directly (Codex, OpenCode, Gemini, Copilot, and every other harness agentsview syncs), incrementally tracks what has already been processed, and writes two memory layers: per-day timeline notes and PARA entity facts. Use when the user says /sessions-memory, 'remember my sessions', 'build memory from sessions', 'what did I work on', 'sync sessions to memory', or wants coding-session his...

Set up a directory as an agent-ready machine context by interviewing the user, recording searchable directories in a YAML manifest, building an indexes/ directory of per-directory index files, and writing an AGENTS.md that tells any agent where to look and how to maintain its own indexes. Use when the user says /setup-agent-machine, "set up this machine for agents", "build an AGENTS.md for my directories", "index my directories for agents", or wants an agent to know where to search on this ma...

Scaffold a MkDocs documentation site with Material theme, initial content structure, and a GitHub Actions workflow to publish to GitHub Pages. Use when the user says /setup-docs-site, "set up docs", "create docs site", "mkdocs setup", or wants to bootstrap a documentation website.

Build or update a Slack channel knowledge base for a given month. Works for any accessible channel given its Slack channel id. Ingests threads via the Slack API and produces temporal monthly memory files.

Collect all Slack conversations a specific person participated in during a date range. Useful for performance reviews, 1:1 prep, or building a picture of someone's contributions. Outputs JSONL via the Slack API.

Go through the Slack threads you participated in, read each one with slackx, judge whether the discussion reached a satisfactory conclusion, and mark resolved threads with a checkmark reaction so future runs skip them. Use this whenever the user wants to triage, review, clean up, or "close out" their open Slack conversations, check which of their threads still need a reply or follow-up, or asks "have all my Slack threads been resolved?" — even if they don't mention slackx by name.

Use the `slackx` CLI to cache and read Slack threads, channel history, DMs, users, and channels from a local SQLite database. It fetches threads via `conversations.replies`, channel history via `conversations.history`, and workspace users/channels, storing everything locally so subsequent reads are instant and incremental. Trigger when the user wants to read, cache, refresh, or poll a Slack thread or channel, render a thread with human-readable author names, search the workspace and cache mat...

Generate a sprint retrospective covering kudos, what went well, what could have gone better, blockers to speed, and action items. Use when the user asks for a sprint retro, retrospective, or sprint review.

Draft and post a structured infrastructure progress announcement to relevant stakeholder Slack channels. Use when the user says /stakeholder-announcement, "announce to stakeholders", "post progress update", or wants to share infrastructure changes (GSAs, WIF, cluster provisioning, batch inference milestones) with stakeholder channels.

Opens the workday by grounding priorities in yesterday's plan and goals, and optionally writes a short start-of-day note. Use when the user says /start-day, start of day, morning planning, or daily kickoff.

Opens the calendar month with a theme, monthly outcomes, and carryover from the prior month review. Use when the user says /start-month, start of month, monthly planning, or new month kickoff.

Opens the calendar week with themes, outcomes, and carryover from last week. Use when the user says /start-week, start of week, weekly planning, or Monday kickoff.

Produces a structured meeting summary document from a transcript file. Use when the user says "summarize meeting", "meeting summary", "summarize transcript", or provides a meeting transcript file to process.

Mark an existing decision as superseded by a newer one, updating status in both records and validating the relationship.

Refresh the indexes of an agent machine context by rebuilding stale or missing index files and indexing directories that appeared under /* manifest paths. Use when the user says /sync-agent-machine, "sync agent machine", "refresh agent indexes", "re-index my directories", or wants agent machine indexes brought up to date.

Bring a batch of blog articles into conformance with the repository writing rules by dispatching one create-article job per article via opencode run, in parallel, then verifying links, front matter, style, and AI disclosure tags. Use when the user says /sync-articles, "sync articles", "update articles according to the rules", "make this year's articles conform", or wants a bulk style and structure pass over article content.

Reconcile the docs/ directory with the codebase so documentation stays current. Detects and fixes drift between code and docs (stale references, missing docs, broken links, nav drift, outdated examples). Use when the user says /sync-documentation, "sync docs", "update docs", "docs are stale", or wants to ensure docs match the current code.

Scan all markdown files in a target directory, extract opinions and the statements that support them, then generate or update an OPINIONS.md file. Use when the user says /sync-opinions, "sync opinions", "extract opinions", or wants to compile opinions from a set of documents.

Ensure the entire repository is consistent by running skills in the right order to sync SDLC, code, tests, documentation, types, and observability. Use when the user says /sync-repository, "sync repo", "keep repo consistent", "run all checks", or wants to detect and fix drift between code, SDLC, and documentation.

Analyze the codebase and reconcile it with the .sdlc/ directory. Creates the .sdlc/ structure if absent, populates context files, creates missing features, updates stale artifacts, and flags drift between code and documentation. With --create-issues, promotes pending (p-prefixed) features to issue-driven ones by creating placeholder GitHub issues.

Build a traceability matrix across a GitHub repository by tracing each issue to its linked PRs and commits and then to the code that implements it, and verifying whether each issue's intents (and acceptance criteria, when present) are still addressed in the current codebase. Use when the user says /trace-issues, "trace issues to code", "build a traceability matrix", "which PRs implemented these issues", "trace the repository", "are the issue intents addressed in the code", or wants a per-issu...

Classify and label a single GitHub issue by type, component, platform, provider, urgency, and importance.

Classify and label incoming GitHub issues by type, component, platform, provider, urgency, and importance.

Proactively scan your open PRs for unresolved reviewer feedback, fan out one analysis agent per PR to write an analysis file per comment, then present decisions and hand off execution to handle-pr-reviewer-feedback. Designed to run on a 10-15 minute schedule so you never have to run handle-pr-reviewer-feedback manually. An orchestrator only: the analysis artifact, vocabulary, and execution all belong to handle-pr-reviewer-feedback.

Update an existing PR description with the minimum changes needed to keep it accurate after new commits have been added.

Update .sdlc/templates/ with the latest canonical best practices, merging intelligently with any user customizations.

Build a profile of users' familiarity with a codebase based on their GitHub contributions, including areas contributed to, issue types resolved, and technologies used.

Generate a pre-vacation (or pre-leave) handoff covering deadline-bound work to land, on-call coverage, in-flight PRs and issues, and items to delegate. Use when the user says vacation handoff, going on leave, OOO handoff, or wants to prepare for being out.

Collect technical assumptions accumulated during design, design and run minimal code experiments to verify the risky ones, and record results before implementation begins.

Visually validate the implemented feature on the current branch BEFORE opening a PR, so the user can confirm the feature works as intended. Captures a CLI demo (via record-asciinema) or a web screenshot/video (via record-playwright) and presents it for human confirmation. For bug fixes, replays the reproduce-issue before-command on the fixed code to produce a comparable after recording. Writes a proof manifest that create-pr consumes, so the recording happens prior to PR creation rather than ...

Judge whether a PR builds the right product. Recovers the customer need behind the linked issue, then checks whether the acceptance criteria and the implemented behavior actually serve that need. Catches "faithfully built the wrong spec", symptom-not-cause fixes, and scope drift. No build or runtime execution (that is verify-pr's conformance role).

Verify that a PR conforms to its specified requirements (acceptance criteria). Combines static criteria-to-code traceability with runtime execution that proves each criterion by building and running the PR. Records CLI demos (asciinema) and web UI captures (Playwright). Does not judge code craft (that is review-pr) or whether the target is the right product (that is validate-pr).

Review notes from the past two weeks to determine what could be demoed.

Guidance for Worktrunk (the `wt` CLI) — git worktree management, hooks, and config. Load when editing .config/wt.toml or ~/.config/worktrunk/config.toml; adding, modifying, or debugging hooks (post-merge, post-start, pre-commit, pre-merge, post-switch, etc.); configuring commit message generation or command aliases; or troubleshooting wt behavior. Also answers general worktrunk/wt questions.

Write a personal status article covering what you have been working on recently and what you currently need. Gathers information from the user through conversation, then produces a polished blog-ready piece.