| name | codex-release-analysis |
| description | Use when evaluating upstream Codex releases, prereleases, app updates, or changelog entries and deciding what Radar evidence Decodex should publish, watch, or absorb from the release. |
Decodex Codex Release Analysis
Use this repo-local skill to turn already-reviewed Radar evidence, official Codex
changelog entries, GitHub release metadata, and GitHub compare metadata into a
source-backed release checkpoint.
Read First
automations/radar/skills/codex-upstream-triage/SKILL.md and
automations/radar/skills/codex-code-analysis/SKILL.md only when routing missing source analysis
Hard Boundaries
- Do not perform fresh upstream source analysis here.
- Do not refresh the upstream review queue here. Treat Radar Review as the shared
upstream evidence producer.
- Do not write
social_candidate/v1, social_post/v1, or
social_publish_reservation/v1. Decodex Publisher owns social artifacts.
- For continuous Radar-derived release and prerelease candidates, consume
upstream_impact/v1 as the shared handoff artifact. Use release deltas, compare
metadata, reviews, signals, and URLs as provenance or gap evidence, not as a second
source-analysis path.
- Use compare metadata to identify PR/commit gaps; route behavior claims that need code
review to
upstream_analysis_required.
Workflow
- Identify the channel: stable release, prerelease, app/mobile changelog, or no new
checkpoint.
- Select the correct comparison:
- stable: current stable -> previous stable
- prerelease: current prerelease -> previous prerelease in the same train
- first prerelease after stable: stable -> first prerelease
- Read existing
upstream_impact/v1 artifacts that match the channel first, then use
release_delta/v1, signal_entry/v1, and upstream_review/v1 artifacts to check
lineage, evidence, and gaps.
- Use
radar backfill-release-range --dry-run when compare PR gaps may need later
source review.
- Choose exactly one primary mode for reporting:
release_pulse: release or changelog headline is the story
watch_note: useful checkpoint with caveats or incomplete analysis
release_rollup: accumulated source-backed Radar evidence explains the changes
delta_explainer: compare metadata maps to existing signals or clear PRs
operator_impact: app-server, plugin, browser, MCP, sandbox, config, auth, or
provider behavior affects external operators
- If the checkpoint has no reader value or only a version tag, write a durable no-op,
defer, skip, or
needs_upstream_analysis decision instead of silent omission.
Output
Return source/timestamp, chosen mode, release-body quality, compare/PR evidence,
matching signal slugs, Control Plane impact if any, Publisher angle if any, and
source gaps.
Promote durable conclusions only through Radar artifacts: shared upstream_impact/v1,
Codex-owned analysis_draft plus radar render-signal, and refreshed
release_delta/v1.