gdd-turn-closeout
Role
You are a deterministic closeout skill. You close the per-turn telemetry gap on runtimes that don't expose a Stop event (codex, gemini, and 11 others). You are a code-level mirror of hooks/gdd-turn-closeout.js (D-10): same conditions, same idempotence, same emitted event shape. The only difference: the JS hook emits the nudge as additionalContext via the harness; this skill prints the nudge directly to the user. A turn is "complete" when the verify command exits 0, the <done> criterion is observable, and the diff stays within the declared scope - any deviation is logged for the run summary.
When to invoke: orchestrator skills ({{command_prefix}}next, {{command_prefix}}design, {{command_prefix}}verify) tail-call this skill as their final step before returning. Adoption is incremental - each orchestrator can wire the tail-call independently; the skill exists as a stable, callable surface today.
Invocation Contract
- Input: none. Operates on
.design/STATE.mdand.design/telemetry/events.jsonlin cwd. - Output: at most one printed line - the nudge - or silent return.
- Latency budget: ≤10ms typical. Read only STATE.md and the tail of events.jsonl.
- Idempotence: if the most recent event is already a
turn_endfor the current(stage, task_progress)tuple, skip append but still print nudge. - Non-blocking: any I/O failure → silent return. This skill never gates the user.
Algorithm
Execute in order; stop at the first early-return.
Step 1 - Read STATE.md
Read .design/STATE.md. Missing or unreadable → return silently (no print, no append).
Step 2 - Parse <position> block
Lightweight-parse only <position>…</position> (regex /<position>([\s\S]*?)<\/position>/ then per-line key: value). Extract stage, status, task_progress. Do not invoke the full STATE parser.
If status != "in_progress" → return silently. Pipeline is initialized / completed / blocked - no turn-end gap to close.
Step 3 - Tail the last event line
Read only the last 8 KiB of .design/telemetry/events.jsonl. Treat as "stale by definition":
- File missing.
- File empty.
- Last line fails JSON parse.
- Last line's
timestampis missing or unparseable.
Otherwise compute now - last_event.timestamp. Gap < 60 seconds → user is actively mid-turn → return silently (next real event closes the gap naturally). Bash one-liner for the tail: tail -n 1 .design/telemetry/events.jsonl 2>/dev/null.
Step 4 - Idempotence check, then append
If last event is already {type:"turn_end", stage:<same>, payload:{task_progress:<same>}} for the exact (stage, task_progress) from Step 2: skip append but proceed to Step 5.
Otherwise append one JSONL line to .design/telemetry/events.jsonl:
json{"type":"turn_end","timestamp":"<ISO 8601 now>","sessionId":"<session-id-or-'turn-closeout'>","stage":"<stage>","payload":{"task_progress":"<N/M>"},"_meta":{"source":"gdd-turn-closeout-skill"}}
Create .design/telemetry/ if missing. Append is a single appendFile-equivalent call (writer assumes append-atomicity per Plan 20-06).
Step 5 - Print the nudge
Match task_progress against ^(\d+)/(\d+)$:
- Numerator equals denominator and denominator > 0 (e.g.
5/5):Stage <stage> complete — run {{command_prefix}}next or {{command_prefix}}reflect. - Otherwise (mid-task, e.g.
3/7,0/0, malformed):Stage <stage> paused mid-task — resume with {{command_prefix}}resume.
One line exactly. No commentary - the nudge is the user-facing surface.
Failure Modes
| Condition | Behavior |
|---|---|
| STATE.md missing/unreadable | Silent return |
<position> absent or malformed | Silent return |
status != "in_progress" | Silent return |
events.jsonl missing | Treat as stale → fall through to append + nudge |
| Last event unparseable | Treat as stale → fall through |
| Last event < 60s old | Silent return |
| Append fails (permission, disk full) | Print nudge anyway; do not surface I/O error |
| Any uncaught throw | Silent return |
Equivalence with the JS hook
This skill and hooks/gdd-turn-closeout.js MUST stay code-level equivalent: same four early-return branches, same 60-second staleness threshold, same idempotence guard (type=turn_end, stage, payload.task_progress), same event shape (only _meta.source differs: gdd-turn-closeout vs gdd-turn-closeout-skill so reflector telemetry can distinguish hook-driven vs skill-driven). Same nudge wording for both N/N and mid-task cases. Change one → change the other in the same plan. Plan 25-09's tests/turn-closeout-hook.test.cjs covers the JS hook.
Non-Goals
- Not a state writer. Never edits STATE.md. The events.jsonl append is the only side effect.
- Not a stage transition.
turn_endis within-stage, not a state-machine move; downstream tools ignore it for transition gating. - Not a Stop-event harness. Cross-runtime Stop-event support at the harness level is out-of-scope for Phase 25 (CONTEXT.md OOS).
Integration Point
Canonical tail-call sites per D-11: {{command_prefix}}next, {{command_prefix}}design, {{command_prefix}}verify. Each orchestrator's final step before returning to the user invokes gdd-turn-closeout. Tail-call wiring intentionally not part of v1.25 (Plan 25-04 ships only the callable surface).
