Persist your session's work so future agents inherit your discoveries. Runs build check (catches bugs before they ship), synthesizes teammate learnings (resolves duplicates/conflicts), then trw_deliver. Use: /trw-deliver
Codex-specific skill: this version is authored for Codex. Follow Codex-native skill and subagent flows, and ignore Claude-only references if any remain.
Run a complete delivery ceremony with pre-flight build verification and post-team learning synthesis. This ensures the codebase is clean and that any duplicate or conflicting learnings from teammate agents are consolidated before delivery artifacts are generated.
Pre-flight build check: Call trw_build_check(scope="full") to run pytest + mypy.
Gate check:
Team learning synthesis (run before delivery):
a. Check for active team: Call task tracking in the current session to check for tasks with team-assigned owners. If no tasks exist or all are owned by the current agent, skip to step 4. (Do not rely on filesystem paths like ~/.claude/teams/ which are CLI-specific.)
b. If a team was active, synthesize learnings:
trw_recall("*", max_results=200) to retrieve all current learnings.trw_status() or the active run's run.yaml).trw_learn_update(learning_id, status="resolved") on each lower-impact duplicate.trw_learn() that includes the highest-impact detail plus an "Alternative finding: ..." note for the differing conclusion. Then call trw_learn_update(learning_id, status="resolved") on all originals in the group.c. If no team was active: skip synthesis, continue to step 4.
Run delivery ceremony: Call trw_deliver() which executes:
trw_reflect — extract learnings from session eventstrw_checkpoint — atomic state snapshottrw_index_sync — update INDEX.md and ROADMAP.md from PRD frontmatterReport summary:
If you catch yourself thinking any of these, stop and follow the process:
| Thought | Why it's wrong | Consequence |
|---|---|---|
| "The build passed earlier, I can skip the pre-flight check" | Build state changes between VALIDATE and DELIVER — new code may have been added | Shipping with a stale build check is the #1 cause of broken deliveries |
| "Team learning synthesis is optional, I'll skip it" | Duplicate/conflicting learnings from teammates pollute future recall | Next session loads contradictory advice — agents get confused and make worse decisions |
| "I'll just call trw_deliver directly, the skill is overkill" | The skill adds build verification + team synthesis that raw trw_deliver skips | Manual delivery skips validation steps — 3x more defects in audits |
trw_build_check + team learning synthesis + trw_deliver into a single workflowtrw_deliver() directly