Changelog

0.2.3

OpenAI Plugins Directory listing metadata. The root and Claude manifests carry extensions.com.openai with explicit interface text: short description "AI security review for PRs" (within the 30-character limit), a long description without plan pricing (the directory forbids pricing in the description), starter prompts, support URL, and the logo path. The description does not name other AI products (directory rule: no references to another AI assistant, model, or platform). Version bump to 0.2.3 across all manifests. No skill behavior change.

0.2.2

Description and README lead with "Aevral is an AI security reviewer for GitHub pull requests." The review scope now lists the live classes (access control, business logic, injection, XSS, SSRF, path traversal, unsafe deserialization, token and session flaws, LLM-integration risks), not only authorization and IDOR. Console wording follows the live site: signing in with GitHub connects the install. No skill behavior change.

0.2.1

Scan API pointer. The skill names AEVRAL_API_KEY as the environment variable to look for (never collected in chat) and links the Scan API docs (https://docs.aevral.com/docs/scan-api.md): start a scan with POST /v1/scans, poll status_url every 30 seconds, stop on any 402. The skill now also loads for a direct "start an Aevral scan" request.

0.2.0

Caps and upgrades. A neutral Check can mean several things (allowance used, payment attention, paid extras paused, hourly or daily cap); the agent reads its title or GET /v1/usage, and shows an upgrade link only when pr.action is not null. With an aevr_ key already in the environment (never collected in chat), the agent may read GET /v1/usage and show the human the billing link an owner or admin opens. The agent never upgrades and never opens the link itself. Prices are quoted only from the public plans endpoint, never from memory (replaces the blanket no-prices rule).

0.1.0

First public portable adapter. Setup skill for the Aevral GitHub App. Reviews run server-side. After App install, the next pull request can already be reviewed, including before console claim.