Trust signals

A public listing in the agent directory carries trust signals so someone deciding whether to use an agent has something to go on.

curl https://api.1claw.co/v1/agents/$AGENT/trust

{
  "badges": ["verified_publisher", "popular"],
  "rating": 4.6,
  "review_count": 23,
  "flagged_for_review": false
}

What is deliberately absent

The public view carries no reviewer notes and no report reasons. Those are queue-internal. Publishing them would publish accusations about an agent's owner, from parties with no accountability.

Reviews

curl -X POST https://api.1claw.co/v1/agents/$AGENT/review \
  -H "Authorization: Bearer $TOKEN" \
  -d '{"rating":5,"comment":"Reliable"}'

Ratings are 1..=5; anything outside is refused. One review per person per agent — reviewing again replaces your previous one. Without that constraint an average rating is whatever the most persistent reviewer wants it to be.

Human users only. An agent rating another agent is a way to manufacture reputation at machine speed.

Reports

curl -X POST https://api.1claw.co/v1/agents/$AGENT/report \
  -H "Authorization: Bearer $TOKEN" -d '{"reason":"spam"}'

One report per person per agent, on the same reasoning as reviews. Reporting again is idempotent rather than an error.

Reports are counted from distinct reporters. Past a threshold a listing is flagged and its badges are withheld pending review — which is why the one-per-person constraint matters: without it a single account could strip every badge off any listing by clicking three times.

A flag is clearable. A moderator dismissal returns the count to zero. A permanent, unappealable state set by a handful of anonymous clicks would be worse than no flag at all.

Endpoints

Method	Path	Auth
GET	/v1/agents/{agent_id}/trust	Public
POST	/v1/agents/{agent_id}/review	Human
POST	/v1/agents/{agent_id}/report	Human