webgrip/ai-skills

Write, amend, and supersede Architecture Decision Records in MADR 4.0.0 — locate or bootstrap an ADR directory with template and registry index in any repo, keep decision history append-only through ratifications, reverts, and supersessions, and enforce file-to-index consistency with a bundled CI-ready validator.

Write, edit, and syndicate technical blog posts: evidence-first drafting, hook workshop, single-dimension edit sweeps with a de-AI-ify pass, fresh-reader testing, and per-platform adaptation for HN, LinkedIn, Substack, dev.to, Reddit, and Lobsters.

Define and maintain a project's ubiquitous language (terms, entities, rules, events) in docs/domain/model.yaml, and generate documentation, diagrams, Context Mapper exports, health scores, and feature specs from it.

PreToolUse hook that hard-blocks plaintext-secret leaks before an edit/write runs: no decrypted secret artifacts, SOPS files must stay ciphertext, and a best-effort gitleaks scan. The Claude Code twin of the estate's opencode guard-secrets plugin — same secrets floor, both tools.

Mine durable, reusable knowledge out of Claude conversation threads with a three-phase workflow — distill each thread into a self-contained digest, consolidate many digests into one deduplicated truth-checked knowledge set, then synthesize repo docs, CLAUDE.md rules, new-skill candidates, and memory updates from it.

Define, groom and facilitate KPI sets for teams and organizations.

Run any ticket board as product owner — Vikunja and ClickUp adapters, research-backed Definition of Ready, agent-ready gating for AI-executed work, flow metrics, and evidence-based closing.

Evaluate an external protocol or standard against a project's architecture seams via parallel research fan-out.

Local skill-usage telemetry: a PostToolUse hook logs every Skill invocation to ~/.claude/skill-usage.jsonl (no network), and the bundled skill reports which skills are used, dead, or undertriggering.

Author, edit, and audit Claude Code skills for token-efficient, high-trigger-accuracy LLM ingestion: a verified token cost model, description-as-router rules, progressive disclosure into reference files, and skill eval methodology.

The whole Webgrip skills estate as one namespaced plugin — every skill loads as webgrip:<skill>, so it can sit alongside another estate's same-named skills.