Perplexica Search

Name/description, SKILL.md, README, and the included Python script all match: the skill queries a local Perplexica instance, returns results with citations, and requires only python3. No unrelated binaries, env vars, or credentials are requested.

SKILL.md restricts the skill to local hosts and instructs running the bundled Python script by full path. The script enforces a hostname whitelist (localhost, 127.0.0.1, ::1, host.docker.internal), rejects redirects, reads an optional config.json from the skill directory, and performs model/provider and search API calls. This scope is appropriate; as a small note, the script’s host-resolution helper is designed to validate allowed hostnames (and checks that host.docker.internal resolves to a loopback/private IP), so confirm the skill's main path validates the base URL before any network calls (truncated code not shown here appears to do so from the rest of the file).

No install spec; the skill is instruction-only with one Python script. This is the lowest-risk install approach and is proportional to the stated functionality.

The skill requests no environment variables, no credentials, and no config paths outside its own directory. That is proportional for a local-only search helper. The only configuration file is an optional config.json in the skill directory (used to set a local perplexica_url).

The skill is not forced always-on and is user-invocable. It does not request system-level persistence or modify other skills. Autonomous invocation is allowed by platform default and is not a concern here.