jedarden
UserCurated Claude Code skills for agent-driven software development
Categories
Indexed Skills (16)
adr
Author a new Architecture Decision Record or review existing ones for quality. Drafts a numbered ADR with steel-manned alternatives and honest consequences, or rates an ADR against a quality checklist and flags strawman alternatives and missing negatives. Use when capturing a design decision, writing an ADR, or auditing existing decision records.
api-design-review
Reviews an API surface — REST/OpenAPI, gRPC/proto, GraphQL schema, or CLI — for design quality and evolvability before it is locked. Rates resource modeling, method semantics, versioning, payloads, and security against an opinionated checklist. Use when an API contract is being designed or about to ship and breaking it later is costly.
diff-review
Language-agnostic structural review of a code diff for correctness bugs and design/cleanup issues, with an adversarial verification pass that suppresses false positives. Collects the diff itself via git, spawns its own reviewers, and emits a severity-ordered report. Use when you want a self-contained, high-confidence review of staged, working-tree, or branch changes before they land.
migration-runbook
Author a reversible cutover/migration runbook — schema migration, data backfill, service cutover, or infra move — that a human or agent can execute step-by-step, with a verification gate and a rollback at every stage. Use when planning or writing a migration, cutover, or backfill that must be safe to abort.
plan-author
Generate a high-quality plan.md from a short brief or idea. Produces the eleven structural categories a plan-review checks for — scope lock, acceptance scenarios, architecture, pre-flight safety, phasing, testing, security, performance, operations, API, and risk — so the draft would pass review on the first pass. Use when starting a new project, port, improvement, or integration and you need a complete plan rather than a blank page.
plan-gap-review
Iteratively review a document for gaps, contradictions, and missing pieces. Spawns agents to brainstorm 20 solutions per gap, picks the best, applies fixes, and re-analyzes. Use when the user wants to review, audit, polish, or find gaps in a plan, spec, or document.
plan-idea-gen
Generate a large pool of ideas (default 100) for a project anchored to its plan.md, then ruthlessly filter to the top K (default 10) via lens-forced generation, clustering, pairwise ranking, and an adversarial kill pass — all inline in a single context (no subagents). Winners become dossiers, ledger entries, and optionally bf beads. Use when the user wants to brainstorm, ideate, expand a roadmap, or asks for "best ideas" / "top N ideas" for a project or plan.
plan-review
Pre-flight review of a software plan that hunts for the decisions an implementer will be forced to make that the plan has not made — and proposes each one — so the plan itself is the decision record and no ADRs are needed mid-build. Runs inline in a single context (no subagents). Produces a verdict (NOT READY / READY AFTER DECISIONS / READY), a decision ledger with a proposed resolution per open fork, an implementer dry-run, a reality check against the real repo and cluster, hard safety caps, and a compact structural sweep. `--lock` writes accepted decisions into the plan where an implementer will actually read them. Use before implementation begins, before decomposing a plan into beads, or whenever a project keeps spawning ADRs after the plan was "done".
postmortem
Author a blameless incident postmortem from an incident description and any available artifacts (logs, timeline notes, chat transcripts), or review an existing draft for blameless tone and analytical depth. Builds a timestamped timeline, quantified impact, contributing-factor analysis, and owned action items. Use when an incident is resolved and needs a written retrospective, or when a draft postmortem needs a quality pass.
readme-review
Reviews a project's README and top-level docs against a quality rubric, judged against the project type (library / CLI / service / app), then optionally redrafts weak sections. Tests whether a stranger can go zero-to-running from the README alone. Use when a README needs a fitness check before release, onboarding, or open-sourcing.
release-readiness
Go/no-go gate run before cutting a release, tag, or deploy. Inspects the repository and everything changed since the last release, rates each release gate against four checklists, and produces an evidence-backed GO / CONDITIONAL / NO-GO verdict. Use when you are about to ship a version and need a final readiness check, or before tagging/deploying a release.
repo-hygiene
Audit a repository for hygiene debt — committed build artifacts, dead GitHub Actions workflows, README version drift, dirty working trees, stash pileups, missing .gitignore coverage, suspicious tracked files — and optionally fix it with one commit per category. Use to check, audit, or clean up repo hygiene.
spec-review
Pre-plan gate that reviews a product or requirements spec (PRD, requirements doc, feature spec) for ambiguity, untestable requirements, and missing non-functional considerations before a plan is authored. Quotes each vague phrase verbatim and proposes a precise rewrite. Use when a spec exists but no plan has been written yet, or before handing a spec to plan-author.
test-plan-review
Suite-level review of an existing test directory or a test-plan document. Finds coverage gaps and tests that will lie to you — what isn't being tested, and which tests pass for the wrong reasons. Use when you have a tests/ tree or a TESTING.md and want to know what's missing before trusting the suite.
threat-model
Produce or review a STRIDE-based threat model for a system. Enumerates assets, trust boundaries, data flows, and entry points, then walks STRIDE per element to emit a threat model document with mitigations — or rates an existing model's coverage and finds gaps. Use when you need a threat model authored from an architecture/design doc or codebase, or when an existing threat model needs a coverage audit.
usage-statusline
Install a live Claude Code statusline showing 5h session and 7d weekly usage against elapsed-time pace, an optional scoped weekly cap, and a rolling count of Claude-co-authored commits. Use when the user wants usage/budget visibility in the prompt line, or asks to set up, fix, or explain the usage statusline.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.