Discover CyberEdge Assets

Use cyberedge-agent as the machine bridge to the CyberEdge gRPC contract. Send exactly one JSON envelope on stdin and parse JSON Lines from stdout. Do not pass human-oriented flags, access the database, invoke scanners, or edit the read-only Web.

Preconditions

• Obtain the human-provided authorization reference and exact targets.
• Refuse ambiguous ownership or authorization.
• Use a fresh request_id and idempotency_key for each mutation. Reuse the same key only when retrying the identical request.
• Send the name and version declared in manifest.json in every invocation context.

Workflow

1. Call CreateScope with normalized target kinds and the authorization reference.
2. Call StartScan with capability scan.passive. Use policy_passive_inventory for broad passive inventory (DNS plus Certificate Transparency), or policy_passive_dns when only direct DNS resolution is requested.
3. Call WatchTask from sequence 0; on reconnect, continue after the last accepted sequence.
4. Stop when the task reaches completed, failed, or canceled.
5. Call GetTaskReport after task.completed to retrieve the deterministic report bundle.
6. Use SearchAudit when the requester needs invocation provenance.
7. Report only facts linked to evidence. Separate errors and coverage gaps from confirmed absence.

For recurring monitoring, call CreateSchedule with the existing scope and passive policy. The minimum interval is 60 seconds. A Schedule never performs discovery itself: each due occurrence creates a normal Task, so follow and report its last_task_id through the same task workflow. Use SearchSchedules to inspect recurrence state and SearchAssetChanges to read evidence-derived appeared or disappeared changes. The first successful run establishes the baseline and produces no changes.

Use snake-case action names. Every envelope includes request_id, idempotency_key, agent_id, skill_name, and skill_version. Example:

{"request_id":"req_...","idempotency_key":"idem_...","agent_id":"agent_...","skill_name":"cyberedge-discover-assets","skill_version":"0.1.0","action":"start_scan","scope_id":"scope_...","policy_id":"policy_passive_inventory"}

Set CYBEREDGE_RPC_SOCKET only when the service does not use /tmp/cyberedge.sock.

Failure handling

• On CAPABILITY_DENIED, stop and report the missing capability; do not try another identity or Skill.
• On IDEMPOTENCY_KEY_REUSED, generate a new key only if the intended input changed.
• On scope validation failure, narrow or correct the target; never bypass validation.
• Retry only when the typed error sets retryable=true.
• Cancel a running task only when the requester explicitly asks or continued execution is unsafe.

Output

Return the scope ID, task ID, terminal state, discovered asset counts by kind, evidence references, coverage limitations, and relevant audit request IDs. Never claim that an unobserved asset does not exist.