<< All versions

Skill v1.0.0

currentLLM-judged scan90/100
vkirill/claude-lane-stack/codex-lane
──Details
PublishedOctober 22, 2026 at 04:32 PM
Content Hashsha256:b8c3b9c04ad4a282...
Git SHA73cdcf8cb118
──Files
Files (1 file, 3.8 KB)
SKILL.md3.8 KBactive
SKILL.md · 115 lines · 3.8 KB

version: "1.0.0" name: codex-lane description: Operate Claude Lane Stack schema-v2 runs from Codex. Use for lane status, start, resume, review, verification, evidence, or acceptance.


Codex Lane

Use the installed typed wrappers under ~/.agents/bin. Keep the run directory, task file, and project root supplied by the user or existing run state. Do not invent a run, task id, path, model, or receipt.

Dependencies

Check the wrappers before acting when the installation is uncertain:

bash
command -v run-controller lane-ctl night-review
run-controller --help
lane-ctl --help

The wrappers may be invoked by absolute path (~/.agents/bin/...) when PATH does not include ~/.agents/bin. Their lifecycle state lives under the run's .agents/runs/<slug>/ directory.

Run and resume

For an existing schema-v2 run, validate and start the durable controller. Keep the run's configured writer by default; add --provider codex only when Codex is the explicitly selected writer:

bash
run-validate --run-dir "$RUN_DIR" --phase pre-dispatch
run-controller start --run-dir "$RUN_DIR" --project-cwd "$PROJECT_CWD"
run-controller watch --run-dir "$RUN_DIR" --timeout 240 --json

run-controller start is idempotent and resumes persisted controller state. Use a fresh bounded watch call after an interrupted session. Read the controller receipt with:

bash
run-controller status --run-dir "$RUN_DIR" --json

Preserve the run's selected native profile, provider, model, reasoning effort, and service tier unless the operator explicitly asks for a change. Do not silently route a run to Codex or override those values.

For one typed lane instead of the DAG controller (use --provider codex only when explicitly selected):

bash
lane-ctl start --run-dir "$RUN_DIR" --task-file "$TASK_FILE" \
--project-cwd "$PROJECT_CWD"

Status and evidence

Read compact state first, then inspect the bounded journal or a specific log:

bash
lane-ctl status --run-dir "$RUN_DIR" --task-id "$TASK_ID" --json
lane-ctl events --run-dir "$RUN_DIR" --task-id "$TASK_ID" --json
lane-ctl tail --run-dir "$RUN_DIR" --task-id "$TASK_ID" --source report --lines 80

Require a trusted complete report.md, matching report_sha256, provider exit 0, and the current attempt before treating a writer as ready for L1. The writer's report is evidence; it is not acceptance.

Review, verify, accept

lane-ctl has no review subcommand. Run the installed read-only review entrypoint when a review is requested:

bash
night-review "$PROJECT_CWD" --project-cwd "$PROJECT_CWD" --run-slug "$RUN_SLUG"

Run the task's immutable verification and ownership checks:

bash
lane-ctl verify --run-dir "$RUN_DIR" --task-file "$TASK_FILE" \
--project-cwd "$PROJECT_CWD"
check-owns-paths "$TASK_FILE" --run-scope

When the task/run requires review, produce the task-specific receipt before acceptance. Use the immutable pre-task commit recorded by the run as BASE_REF; if it is missing, report the missing baseline. A repository-wide night review does not substitute for this receipt:

bash
night-review-engine review-task --repo "$PROJECT_CWD" --run-dir "$RUN_DIR" \
--task-file "$TASK_FILE" --base-ref "$BASE_REF" \
--output "$RUN_DIR/artifacts/$TASK_ID/review.json"

Finally apply the typed acceptance gate:

bash
lane-ctl accept --run-dir "$RUN_DIR" --task-file "$TASK_FILE" \
--project-cwd "$PROJECT_CWD"

Acceptance must produce artifacts/<task-id>/acceptance.json and a current accepted state. If any receipt is stale or missing, stop and report the exact gate failure instead of editing control-plane artifacts.

Do not run provider commands directly, rewrite report.md, bypass lane-ctl, or claim browser or user-specific verification from a fixture. This plugin does not replace native Codex compaction or add an MCP server.

All versions