quickcall-dev/skills

Agent Skills for documentation management and fleet orchestration. Run parallel AI workers with DAG ordering, git worktree isolation, reviewer gates, and autonomous research loops.

Works with Claude Code, Cursor, GitHub Copilot, Gemini CLI, OpenAI Codex, Goose, Roo Code, JetBrains Junie, and 25+ other agents.

Install

# All skills
npx skills add quickcall-dev/skills

# Just one
npx skills add quickcall-dev/skills --skill dag-fleet

Skills

Documentation

doc — structured documentation management

Create and manage structured documentation — experiments, plans, findings, checkpoints, research, learnings. Config-driven, parallel-safe.

Use when: starting new investigations, tracking experiment progress, writing plans, recording findings, or creating checkpoints at natural stopping points.

Commands: start, expt, plan, finding, ckpt, research, review, learn, list, status, resume

npx skills add quickcall-dev/skills --skill doc

Fleet Orchestration

fleet-plan — analyze tasks and generate fleet configs

Analyze a task, pick the right fleet type, and generate a ready-to-launch fleet (fleet.json + prompt.md files). Discovers available fleet skills dynamically.

Use when: you want to run work in parallel and need help choosing between dag-fleet, worktree-fleet, iterative-fleet, or autoresearch-fleet.

npx skills add quickcall-dev/skills --skill fleet-plan

dag-fleet — DAG-ordered parallel workers

Persistent, budgeted, DAG-ordered runner for parallel claude -p or codex exec workers in tmux. Supports dependency ordering, per-worker budget caps, mixed models/providers, and concurrency limits.

Use when: you need persistence across sessions, per-worker budget caps, dependency ordering, or mixed models/providers per worker.

Commands: launch, status, kill, report, relaunch-worker, feed, view

npx skills add quickcall-dev/skills --skill dag-fleet

worktree-fleet — git-worktree-isolated parallel workers

Independence-validated parallel fleet that runs each worker in its own git worktree. Each worker gets its own branch for merge-safe isolation.

Use when: tasks touch non-overlapping files and you need merge-safe isolation with each worker on its own branch.

Commands: launch, status, merge, cleanup

npx skills add quickcall-dev/skills --skill worktree-fleet

iterative-fleet — reviewer-gated iterative cycles

Reviewer-gated iterative fleet for headless workers that run in cycles until a designated reviewer approves the output. A reviewer worker reads all worker logs, writes a verdict (lgtm | iterate | escalate), and the orchestrator decides whether to continue, pause, or stop.

Use when: work needs multiple rounds of iteration with a quality gate. Never kills or restarts workers automatically — the operator owns all decisions.

Commands: launch, status, pause, resume, kill

npx skills add quickcall-dev/skills --skill iterative-fleet

autoresearch-fleet — autonomous research loop

Karpathy-inspired autonomous research loop. Agent edits one file, evals, keeps or discards, repeats. Plateau-triggered web search breaks through ceilings. Git as state machine.

Use when: you need autonomous, iterative optimization with automatic evaluation. Runs until stopped or budget exhausted.

Commands: launch, status, view, report, pause, resume, kill

npx skills add quickcall-dev/skills --skill autoresearch-fleet

How It Works

Each skill follows the Agent Skills open standard — a SKILL.md file with YAML frontmatter and markdown instructions. Agents load the skill when it matches the task at hand. Fleet skills bundle bash scripts that orchestrate parallel workers in tmux sessions.

skills/
├── doc/                    # Documentation management
│   ├── SKILL.md
│   ├── config/defaults.yaml
│   ├── scripts/            # 13 command scripts
│   └── references/
├── dag-fleet/              # DAG-ordered parallel workers
│   ├── SKILL.md
│   ├── scripts/            # 7 orchestration scripts
│   ├── lib/                # Shared fleet libraries
│   └── references/
├── worktree-fleet/         # Git-worktree isolated workers
├── iterative-fleet/        # Reviewer-gated cycles
├── autoresearch-fleet/     # Autonomous research loop
└── fleet-plan/             # Fleet config generator

Testing

Tests live in test/ and are not part of the skills themselves. They validate skill wiring without calling real APIs.

# Run all test suites (from repo root)
bash test/fleet/dag-fleet/fixtures-claude/run-all.sh skills/dag-fleet
bash test/fleet/dag-fleet/fixtures-codex/run-all.sh skills/dag-fleet
bash test/fleet/worktree-fleet/fixtures-claude/run-all.sh skills/worktree-fleet
bash test/fleet/worktree-fleet/fixtures-codex/run-all.sh skills/worktree-fleet
bash test/fleet/iterative-fleet/fixtures-claude/run-all.sh skills/iterative-fleet
bash test/fleet/iterative-fleet/fixtures-codex/run-all.sh skills/iterative-fleet
bash test/fleet/autoresearch-fleet/run-all.sh skills/autoresearch-fleet

59/59 scenarios passing across all suites.

License

Apache-2.0 — see LICENSE