Agent platforms

The canonical architecture is:

Natural-language request
  -> platform Skill
  -> installed shodan-skill CLI
  -> shared validation, safety, transport, output, and redaction
  -> Shodan services

Install the Python CLI first, then install one generated Skill bundle into the platform discovery layout:

python -m pip install shodan-skill
python scripts/install_skill.py --platform codex
python scripts/install_skill.py --platform openclaw
python scripts/install_skill.py --platform claude-code
python scripts/install_skill.py --platform hermes

The repository-root SKILL.md is the canonical entry for community directories. The files under platforms/ are generated from that Skill and focused references; do not edit or submit them as independent Skills. The installer prompts before replacing an existing bundle. Use --yes only when replacement is intended.

Request behavior

An explicit natural-language operation request is handled like an explicit CLI command under the configured safety mode. The agent must validate the stated targets, preserve scope, and use the installed CLI. A configured key, available credits, or Enterprise entitlement does not authorize an unstated operation.

Examples:

Show the current Shodan record for 8.8.8.8.
Count hosts matching port:443 without running a full host search.
Dry-run a Shodan scan request for the target I explicitly authorized.
Stream ten records for ports 22 and 443.

The Agent Skill uses progressive references for search, scans, streaming, Enterprise services, schemas, and safety. User-facing workflows belong in this manual; the platform bundle remains concise and operational.