Skip to content

v2.1.202

Choose a tag to compare

View all tags
@mike1858 mike1858 released this 06 Jul 21:21
v2.1.202
1398a62

2.1.202

(☝️ Click to see full diff.)

+3,217 tokens

  • NEW: Agent Prompt: /code-review part 2 low effort minimum findings mode — Adds a low-effort /code-review mode that reads the diff once, skips test and fixture hunks, avoids subagents and full-file reads, and targets hunk-visible runtime-correctness findings with one extra pass before returning (none).
  • NEW: Data: Governed GitHub CLI shim header and routing — Adds the per-session governed gh shim text that routes github.com requests without customer credentials through the agent proxy while letting customer-token and GitHub Enterprise invocations use the real gh, including host/repo/origin detection, proxy/CA setup, and proxy-injected tokens.
  • Agent Prompt: /review slash command — Replaces the medium-effort JSON-findings review flow with a concise, sectioned PR review covering overview, code quality/style, improvement suggestions, risks, correctness, project conventions, performance, tests, and security.
  • Agent Prompt: Security monitor for autonomous agent actions (second part) — Clarifies that unknown repository visibility is not itself a blocking reason for data exfiltration or out-of-place publication checks, while keeping content sensitivity and same-session remote repoints as separate risk signals.
  • Data: Claude Code live documentation sources; Data: Claude Tag (Claude in Slack) reference; and Skill: Claude Code configuration guide — Clarify that .md Claude docs URLs are for fetching only and user-facing links should drop the trailing .md so they open the rendered docs page.
  • Skill: Dynamic pacing loop execution; Skill: /loop self-pacing mode; System Prompt: Monitor fallback heartbeat guidance; and Tool Description: Snooze (delay and reason guidance) — Make loop re-arming an explicit per-turn decision, handle task notifications before deciding whether to continue, and end loops by calling the wakeup tool with stop: true instead of omitting the wakeup call.
  • Skill: PR explainer — Requires PR walkthrough artifacts to answer what problem the PR solves, why it matters, how the PR solves it, what alternatives were considered, and why the chosen approach is better, or state plainly when the PR materials do not provide that evidence.
Assets 2
Loading