Conductor for the dev pipeline: detects a task's tier (bug / feature / architecture / client), prints the exact chain of sub-skills it will run and skip with reasons, confirms, then drives them in order. Matches ceremony to stakes.
0 on GitHubby Matheus Malagrisv0.8.0MITupdated 6 days agoengineering-workflow
/
Plugin dev-flow from marketplace agent-skills in mqmalagris/agent-skills. Commands read from the repo's manifest — 6 days ago.
What it installs
Skills (1)
dev-flow
Route a dev task through the right subset of the canonical skill chain (adhd → grill-me → to-prd → compass → blueprint → code → review-pass → pr-craft), matching ceremony to stakes. Detects the task tier (bug / feature / architecture / design / client), shows the exact chain it will run and what it
Estimate how long a dev task will really take when it's built AI-assisted (an agent writes the code, a human reviews, tests, ships). Grounds every number in the actual code first, counts agent work in tool-call rounds, then adds the human layer (diff review, local run, browser QA, PR/CI, review feedback, PM back-and-forth) as separate lines. Use this whenever someone asks for an estimate, "how long would X take", "how many hours", sizing, scoping, a quote for a PM or client, or a comparison of options by effort — even when they don't say "estimate" and even for a quick ballpark. Also use it before replying to a ticket, Lobby/Teamwork/Slack thread or PR comment that asks "estimate to complete?". Adapted from ZhangHanDong/agent-estimation (MIT).
Render audit findings as a designed, paginated A4 PDF report with a cover page, severity donut and category bar charts, colored severity chips, evidence-backed strengths, prioritized recommendations, and copy-ready GitHub issue blocks.
Turn a settled scope into an implementation plan at docs/plans/NNNN-<slug>.md: files affected, phased sequence gated on numbered acceptance examples, edge-case ledger, rollback and risk. Consumes PRDs + ADRs. Formerly heist.
Maintain a monthly brag document and roll it up into a promo packet, self-review, CV material or interview story seeds. Lucas Faria's 4-field template + Julia Evans' impact-first categories, with ownership, status, evidence and story seeds recorded each month. Feeds cv-craft.
Worktrunk is a CLI for Git worktree management, designed for parallel AI agent workflows. This plugin provides configuration guidance (LLM commit messages, project hooks, worktree paths), automatic activity tracking (🤖/💬 indicators in `wt list` showing active Claude sessions), and a `/wt-diff` pane showing each worktree's changes.
Multi-agent orchestration grounded in an LLM wiki. /dev-loop:orchestrate splits one goal into a dependency graph of parallel Claude Code sessions, each in its own worktree, and supervises them natively on Orca (a tracked Task + Dispatch per phase, pushed worker_done / question / escalation events instead of polling, stall detection; raw tmux fallback). Every plan decision must cite a page in a bundled best-practice wiki of 400+ pages across 10 domains, checked by machine gates and a fresh-context plan reviewer. Insights from your sessions are queued and turned into reviewed wiki PRs when you run /dev-loop:knowledge-flush (automatic flushing is opt-in and never bypasses permission checks).
This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.