@codex-agent/cli 0.1.0-main.11.sha41f680f → 0.1.0-main.13.sha3ad2ccf

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -5,15 +5,17 @@ Command-line diagnostics and project bootstrap helpers for the [codex-agent](htt
5
5
  Run the latest published version without a global installation:
6
6
 
7
7
  ```bash
8
- npx --yes @codex-agent/cli@latest init --json
8
+ npx --yes @codex-agent/cli@latest context init --json
9
+ npx --yes @codex-agent/cli@latest context refresh --json
9
10
  npx --yes @codex-agent/cli@latest doctor --json
10
11
  npx --yes @codex-agent/cli@latest context save --proposal context-proposal.json --json
12
+ npx --yes @codex-agent/cli@latest context lint --json
11
13
  npx --yes @codex-agent/cli@latest migrate navigation --from /path/to/project --json
12
14
  npx --yes @codex-agent/cli@latest eval --json
13
15
  ```
14
16
 
15
- Run these commands from the target repository. Use `npx @codex-agent/cli@latest help` to list every command. Initialization and context saving preview changes by default; pass `--apply` only after reviewing the result. Existing context updates also require `--update`.
17
+ Run these commands from the target repository. Use `npx @codex-agent/cli@latest help` to list every command. Context initialization, refresh, and saving preview changes by default. For init or refresh, pass `--apply --plan-hash <reviewed-plan-hash>` so repository drift cannot change the reviewed proposal; context saving uses `--apply`, and existing curated-context updates also require `--update`.
16
18
 
17
19
  `migrate navigation` discovers navigation-based Markdown context trees, skips incompatible runtime material by default, and writes native indexed context only with `--apply`.
18
20
 
19
- `eval` validates focused positive, negative, and overlap skill-routing fixtures plus required and forbidden behavior contracts for every bundled skill and canonical agent. It does not run end-to-end or A/B prompt comparisons.
21
+ `context lint` is read-only and checks catalog lifecycle, provenance hashes, review dates, duplicates, and orphan documents; `--strict` makes warnings fail. `eval` validates focused positive, negative, and overlap skill-routing fixtures plus required and forbidden behavior contracts for every bundled skill and canonical agent. The source workspace additionally provides opt-in model execution through `npm run eval:model`; it is not part of the published CLI or offline gate.