check-agent-annotations
What this skill does
- •Reads the open annotation inbox:
.agent-annotations/inbox.jsonl - •Optionally reads the resolved archive:
.agent-annotations/inbox-resolved.jsonl - •Summarizes annotations (route, locator hints, comment, attachments)
- •Produces an actionable plan (what to search, where to edit, how to validate)
Where annotations live
Open (active):
- •
.agent-annotations/inbox.jsonl - •
.agent-annotations/assets/open/
Resolved (archive):
- •
.agent-annotations/inbox-resolved.jsonl - •
.agent-annotations/assets/resolved/
If files don’t exist, treat as “no annotations yet”.
How to use locators
Given element.primary:
- •
testid: search repo for the value (e.g.data-testid="...") - •
id: search forid="..." - •otherwise: use
textHint,role, and alternate locators to narrow down components/files
Browser validation (only if user asks)
Only do browser validation if the user explicitly requests it. If a browser automation tool is available (browser skill, MCP tool, Playwright, etc.), you may:
- •open the annotated URL
- •locate the element using
data-testid/ role / text hints - •capture before/after screenshots Otherwise, provide manual validation steps.
Marking items resolved
Preferred method (if the receiver is running):
- •Read config (optional) and token:
- •
.agent-annotations/config.jsoncontainsreceiverBaseUrl(defaults tohttp://localhost:8787) - •
.agent-annotations/token.txtcontains the token
- •POST status update:
bash
BASE_URL="$(node -p "try{require('./.agent-annotations/config.json').receiverBaseUrl}catch(e){'http://localhost:8787'}")"
TOKEN="$(cat .agent-annotations/token.txt)"
curl -s -X POST \
-H "Content-Type: application/json" \
-H "X-Annotation-Token: $TOKEN" \
-d '{"status":"resolved"}' \
"$BASE_URL/annotations/<ID>/status"
Fallback (if receiver is not running):
- •move the record from
inbox.jsonltoinbox-resolved.jsonland setstatusto"resolved".
Reference
See references/ANNOTATIONS_SCHEMA.md.