By Dev-Jahn
SSOT-anchored agentic development workflow harness: one-click project setup, task registry with a global naming convention, zero-token roadmap rendering, round closeout, external review ingestion, and scoped SSOT audits.
This skill should be used when the user runs "/jahns-workflow:audit", asks to "audit the SSOT", "verify the design doc", "check the spec for errors", or after a bulk SSOT edit triggers the quarantine rule. Runs a scoped, tiered independent verification of the SSOT — verification stays cheap (seconds, not hours); it never runs production test suites or expensive builds.
This skill should be used when the user runs "/jahns-workflow:init", asks to "initialize the workflow harness", "set up jahns-workflow", "adopt the workflow in this project", or "re-sync the workflow setup". One-click setup for new projects and non-destructive retrofit for projects already in progress.
This skill should be used when the user runs "/jahns-workflow:review", pastes an external review reply (e.g. from web ChatGPT / GPT reviewer) to be processed, or asks to "ingest the review", "process the reviewer feedback", "record the external review". Preserves the review verbatim and triages findings into the task registry.
This skill should be used when the user runs "/jahns-workflow:round", says to "close the round", "wrap up this round", "finish the work cycle", or when an autonomous work round (implement → verify → push) reaches its end and the project CLAUDE.md mandates round closeout. Updates the task registry, PROGRESS, roadmap, SSOT views, and produces the external review packet.
This skill should be used when the user runs "/jahns-workflow:status", asks "what's the status across my projects", "show the project dashboard", "which tasks are active/blocked", or wants a cross-project overview of branches, rounds, and task progress.
Modifies files
Hook triggers on file write and edit operations
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
A Claude Code plugin that harnesses SSOT-anchored agentic development: one global naming convention across all projects, a machine-validated task registry, zero-token roadmap/digest generation, round-based progress discipline, external-review ingestion, and scoped SSOT audits.
Generalizes a research-workflow battle-tested on theory-heavy repos, but applies to any programming project (web, systems, ML) that anchors work on one design document.
<type>/<kebab-slug> (feat|fix|perf|gate|spike|decision|docs|chore) registered in tasks.yaml with explanatory titles before first use. Bare codenames (P0, E3, Q1) are rejected by a validator hook.blocker > major > minor on review/audit findings.tasks.yaml (registry), ROADMAP.md (generated view), PROGRESS.md (log, auto-archived monthly), ADRs (decisions), reviews dir (external feedback, verbatim).Full convention: references/conventions.md.
/jahns-workflow:init | One-click setup; non-destructive retrofit for in-progress projects (incl. agent-memory cleanup, project registration) |
/jahns-workflow:round | Close a work round: sync registry → PROGRESS entry + archive → refresh views → review-request packet |
/jahns-workflow:review | Ingest an external review reply: preserve verbatim → verify each finding → register as tasks |
/jahns-workflow:audit | Scoped, tiered SSOT audit (consistency / independent re-derivation / cheap-oracle checks). Cost cap is wall-clock, not hardware — never runs production suites or expensive builds |
/jahns-workflow:status | Cross-project terminal dashboard (branches, rounds, active/blocked tasks). Registry entries can be local (path) or remote (repo: owner/name, fetched via gh api — for projects not cloned on this machine) |
spec-auditor agent | Independent verifier fanned out by the audit skill |
| SessionStart hook | Injects digest + active tasks on startup/resume/clear/compact (capped ~8KB; no-ops in ~30ms in non-initialized projects) |
| PostToolUse hook | On tasks.yaml edits only: schema validation (exit-2 feedback) + deterministic ROADMAP.md regeneration. ~13ms no-op otherwise |
All rendering/validation is plain Python (scripts/jw_*.py, run via uv) — zero LLM tokens.
git, bash, uv on PATH (scripts use PEP 723 inline deps; first run downloads pyyaml once).# from a marketplace that lists it
/plugin install jahns-workflow
# or for local development
claude --plugin-dir ~/workspace/jahns-workflow
Then in each project: /jahns-workflow:init. Restart Claude Code after install so hooks load.
.jahns-workflow.yml # config: SSOT path, dir mapping, oracles, audit watermark
tasks.yaml # THE codename registry (validated on every edit)
ROADMAP.md # generated Mermaid dependency graph + task table (GitHub renders it)
PROGRESS.md # round log (older months auto-archived to docs/progress/)
docs/CONVENTIONS.md # verbatim copy of the global convention
docs/ssot/ # generated: sections/ split, INDEX.md, DIGEST.md
docs/adr/ docs/reviews/ # decisions; review request/feedback/audit records
CLAUDE.md # gains a marker-delimited workflow stanza
Skills load only on invocation (descriptions ~60 words each); hooks are command-type (no LLM); the only recurring context cost is the SessionStart injection (digest capped at 150 lines, task summary at 8 lines per state) — which replaces re-reading a 60–150KB SSOT after every compaction.
npx claudepluginhub dev-jahn/jahns-cc-marketplace --plugin jahns-workflowMarkdown link graph and staleness detection for Claude Code. Automatically tracks cross-references between markdown documents and detects when linked content becomes stale.
Karpathy-style autonomous LLM research loop adapted for generic ML codebases. Ships two skills — setup and run — plus an `ar` helper CLI that owns launch orchestration, metric extraction, atomic checkpoint commit, and chain-mode transitions so the main Claude session stays context-light across hundreds or thousands of iterations. v0.3.0 drops the wandb/accelerate monoculture: metric backend is pluggable (wandb / tensorboard / log-scan / custom with auto-detection), distributed-framework resolution is auto-detected (accelerate/deepspeed/fsdp/ddp/lightning/none), a new `hydra` entry pattern renders a Hydra-override-style train wrapper, and `--checkpoint-glob` gives priority-0 control over checkpoint discovery for Lightning / plain-torch / custom layouts.
Harness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses
Feature development with code-architect/explorer/reviewer agents, CLAUDE.md audit and session learnings, and Agent Skills creation with eval benchmarking from Anthropic.
Reliable automation, in-depth debugging, and performance analysis in Chrome using Chrome DevTools and Puppeteer
Comprehensive feature development workflow with specialized agents for codebase exploration, architecture design, and quality review
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques
Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.