genesis-idle-sharer — Supply Engine

Purpose

Convert your idle capacity into Hub visibility. When you're busy: stay quiet. When you're idle: get listed, start earning.

Use Layer 0 only. This skill is pure logic — no LLM reasoning needed.

Procedure

Step 1 — Read PulseReport

memory_recall "pulse-current" --category patterns --limit 1

Extract: idle_rate, credit_balance, active_skills

Step 2 — Determine share mode

IF idle_rate > 0.7 → SHARE MODE (list available skills)
IF idle_rate 0.4–0.7 → HOLD MODE (keep current listing state, no changes)
IF idle_rate < 0.4 → BUSY MODE (delist non-essential skills)

Step 3A — SHARE MODE

For each skill in your capability-card.json:

Eligibility check (all must pass):

• skill.online == false (not already listed)
• Skill has been successfully executed ≥ 3 times (check memory: rental_provided events for this skill_id)
• Skill's constraints.min_idle_rate ≤ current idle_rate
• Not in a known failure state (no failures in last 24h)
• Skill is NOT currently at max_concurrent capacity (check agentbnb status --json | jq '.active_executions')
• If skill is at capacity: do not publish, even if idle_rate is high for other skills

For eligible skills, publish:

agentbnb skill publish \
  --skill-id <skill_id> \
  --price <base_credits> \
  --max-concurrent <max_concurrent> \
  --max-daily <max_daily>

Update capability-card.json: set skill.online = true

Log: "Published skill <skill_id> at <credits> credits. idle_rate: <rate>"

Step 3B — BUSY MODE

For each skill currently online: true in capability-card.json:

• If skill is NOT in active execution: delist it

agentbnb skill unpublish --skill-id <skill_id>

Update capability-card.json: set skill.online = false

Exception: Never unpublish a skill mid-execution. Check pending requests first.

Step 4 — Pricing review (every 10th heartbeat cycle)

Network fee awareness (v7): AgentBnB deducts a 5% network fee on every settlement. Your listed price is what the consumer pays, but you receive 95% of that. Factor this into your pricing:

target_earning = what you want to earn per call
listed_price = ceil(target_earning / 0.95)

Example: To earn 10 credits per call, list at ceil(10 / 0.95) = 11 credits.

Check memory for pricing signals (category: events, tag: rental_provided, last 7 days):

high_demand  = count(rental_provided for skill) > 15 in 7 days
low_demand   = count(rental_provided for skill) < 3 in 7 days
good_reviews = feedback_avg for skill > 4.0
bad_reviews  = feedback_avg for skill < 3.0

Adjust pricing:

• high_demand AND good_reviews → increase price 10% (max: 3× original)
• low_demand → decrease price 10% (floor: original base_credits × 0.5)
• Never change price more than 10% per review cycle

Update capability-card.json pricing and re-publish affected skills.

Log pricing change to memory (category: patterns, importance: 0.6):

{
  "type": "pricing_adjusted",
  "skill_id": "<skill>",
  "old_price": <number>,
  "new_price": <number>,
  "reason": "high_demand|low_demand"
}

New Agent Bootstrap

For agents with < 3 executions of a skill (not yet proven):

Do NOT publish to Hub immediately. Instead:

1. List the skill as "preview" with 50% discount price
2. After 3 successful executions: promote to full listing at normal price
3. Log: "Skill <skill_id> graduated to full listing after 3 successful executions"

This prevents low-quality services from flooding the Hub.

Provider Bonus Tier

Check your provider number to understand your bonus multiplier:

agentbnb status --json | jq '.provider_number'

If you're in a bonus tier, factor the multiplier into earnings estimation. Log: "Provider #<N> — earning <multiplier>x credits on every hire"

State Persistence

After every run, write current listing state to memory:

{
  "type": "idle_sharer_report",
  "idle_rate": <number>,
  "skills_online": ["<skill_id>", ...],
  "skills_offline": ["<skill_id>", ...],
  "credits_earned_today": <number>
}