v1.2.0 -- Agent Onboarding, Cost Estimation & Usage Metadata

What's New

Agent Onboarding (scripts/onboard.py)

A setup wizard that adapts to its audience:

# AI agents get a JSON capabilities manifest
uv run scripts/onboard.py --agent

# Humans get an interactive setup interview
uv run scripts/onboard.py --interactive

# Quick config status check
uv run scripts/onboard.py --check

The --agent manifest includes command inventory, decision trees, output contracts, and configuration status -- everything an AI agent needs to integrate the skill.

AGENTS.md

A structured briefing file optimized for agent consumption: capabilities table, 4 decision trees, copy-pasteable workflows, output JSON schemas for every command, error recovery guide, and pricing reference.

Cost Estimation (--dry-run)

Preview estimated costs before running research:

uv run scripts/research.py start "Analyze security architecture" \
  --context ./src --dry-run

Outputs a JSON cost estimate with context upload costs, research query costs, and total. Estimates are heuristic-based (the Gemini API does not return token counts) and clearly labeled as such.

Post-Run Usage Metadata

After research completes with --output-dir, metadata.json now includes a usage key:

{
  "usage": {
    "disclaimer": "Estimates based on output size and pricing heuristics.",
    "output_bytes": 45000,
    "estimated_output_tokens": 11250,
    "estimated_input_tokens": 250000,
    "estimated_cost_usd": 1.22,
    "context_files_uploaded": 42,
    "source_urls_found": 15
  }
}

The compact stdout JSON also includes estimated_cost_usd and duration_seconds.

Files

New	Purpose
scripts/onboard.py	Setup wizard + capabilities manifest
AGENTS.md	Structured agent briefing

Modified	Changes
scripts/research.py	--dry-run, pricing constants, usage metadata
SKILL.md	"For AI Agents" section, version bump
README.md	Onboarding and cost estimation sections

See CHANGELOG.md for full details.