git:20261005.ea10fe1 to git:20261012.11f465d

7 added, 10 removed. Audit A to A.

---

name: context-guru-insights

- description: Analyze measured context, token, prompt-cache, idle-time, and tool costs. Use when asked what wastes tokens or money, what to optimize, or for context-guru recommendations.

+ description: Moved to the status skill. Kept for one release so existing habits still work.

---

- # Analyze Codex context cost

+ This command moved to `context-guru-status`. Say so in one line, then run the new one from this skill directory and print its output exactly as it is in one code block:

- Run `python3 ../../scripts/insights.py all`. Report its ranked findings in order and preserve every

- measurement label and fix command exactly. The script—not the model—does the arithmetic and sorting.

- Stop on `not_installed` or `proxy_unavailable`. On a subscription plan, describe dollar values as

- estimated usage-limit value rather than a lower bill.

+ ```bash

+ python3 ../../scripts/cg.py home

+ ```

- Codex command sandboxes may block the local `127.0.0.1` dashboard. On `proxy_unavailable`, retry

- the same read-only command with `sandbox_permissions="require_escalated"`, explaining that

- loopback access is needed. If that retry cannot run, report the measurements as unverified rather

- than claiming that the proxy is unavailable.

+ If loopback access is blocked, retry the same read-only command with `sandbox_permissions="require_escalated"`.

+ If it still cannot run, say proxy state is unverified rather than guessing.