feat(shift_report): end-of-day shift report (Phase 2 Step 7) by AVADSA25 · Pull Request #12 · AVADSA25/codec

AVADSA25 · 2026-05-02T18:31:20Z

Summary

Phase 2 Step 7 — final Phase 2 step. End-of-day summary of everything CODEC observed and accomplished. Single PWA notification with type="shift_report" and a 5-section markdown body the dashboard renders inline. Phase 2 is complete after this lands.

Three trigger paths

Trigger	When	Detected by
time	Wall clock matches `config.shift_report.daily_at_hour:daily_at_minute` (default 18:00 local)	`codec-observer` daemon loop
idle	`CGEventSourceSecondsSinceLastEventType` ≥ `config.shift_report.idle_minutes * 60` (default 30 min)	Same
manual	User invokes via skill name (`"shift report"`, `"what did i do today"`, `"summarize my day"`)	Chat / voice / MCP — bypasses dedup

Per-day dedup at ~/.codec/shift_report_state.json ensures time + idle fire at most once per local date; whichever wins suppresses the other. Manual invocations always work (so you can re-summarize any time).

5 sections, ~500-1500 words

Completed tasks — successful tool_result / crew_complete / hook_fired / trigger_fired counts + most-fired tools
Blocked / stuck moments — stuck_warning / step_budget_exhausted / ask_user_question_timeout / trigger_blocked
Observed work patterns — app time-share from observation_tick METADATA + persisted observer summary count
Pending decisions — open type="question" notifications + unreviewed skill proposals
Tomorrow's open threads — crew_start without matching crew_complete (incomplete operations)

Each section gracefully falls back to a "(no data)" placeholder for quiet days.

Commits

sha	what
`1601c4d`	Implementation: audit constants + skills/shift_report.py + codec_observer.py integration + 20 tests + AGENTS.md

(Single commit — Step 7 is contained enough that a separate design doc + impl split would be overhead. The blueprint already covered §0-§10 design; this PR description serves as the design summary.)

Test plan

pytest tests/test_shift_report.py → 20/20 passing in 0.16s
All filesystem paths redirected to tmp_path per test — NO real ~/.codec/* writes
All audit emits redirected to tmp_path — NO real audit.log writes
State files clean post-test: pending_questions=0, Apple Reminders=0, /tmp/codec_*.txt=0, shift_report_state.json absent
Per Step 4 contract: did NOT run full pytest suite

Test breakdown (20)

Area	Count	What
Assembly	8	Empty inputs / completed / blocked / observed_patterns / pending decisions / proposals / tomorrow / observer summaries
Notification + state	3	post type=shift_report / per-day dedup / manual bypasses dedup
Kill switch + config	3	disabled / default enabled / config overrides honored
Audit emit	3	started+completed share cid / extras carried / kill-switch silent
Observer integration	3	idle path fires / off-window doesn't / dedup honored across paths

§6 audit events (2 new)

Event	Source	level	When
`shift_report_started`	`codec-shift-report`	info	Assembly opens — `extra.trigger_kind`
`shift_report_completed`	`codec-shift-report`	info	Assembly closes with summary stats — `extra.{trigger_kind, sections_included, word_count, audit_records_scanned, notifications_scanned, observer_summaries_used}` + top-level `duration_ms`

Both share a single correlation_id (multi-emit op per Step 1 §1.4).

What this PR does NOT do

No new PM2 service — runs inline in codec-observer's daemon loop
No Apple Reminders / Calendar / Notes
No _HTTP_BLOCKED change
No modification of codec_self_improve.py (Step 4 work, sealed)
No auto-fire at merge time — observer fires only when time or idle conditions match (default 18:00 OR 30 min idle)
No new PWA endpoint — the existing notifications endpoint surfaces the report; type="shift_report" is the rendering hint

Kill switch

SHIFT_REPORT_ENABLED=false env var blocks all three trigger paths (time / idle / manual). The skill is still discoverable but run() returns the disabled message immediately.

After merge

git pull on main checkout
cp ~/codec-repo/skills/shift_report.py ~/.codec/skills/shift_report.py — copies the skill to the user runtime install (matches Phase 1 Step 4 pattern)
pm2 restart codec-observer — picks up the integration
To test manually: in chat or via MCP, "shift report" → posts a notification (bypasses dedup so you can iterate)
To verify time/idle paths: wait for 18:00 local OR be idle 30 min OR temporarily set daily_at_hour to a near-time value in ~/.codec/config.json:shift_report and pm2 restart codec-observer

Phase 2 status after this PR

✅ Step 5 (Continuous Observation Loop) — merged via PR feat(observer): continuous observation loop (Phase 2 Step 5) #9 + hotfix fix(observer): ocr_enabled flag — bypass screencapture popup storm #10
✅ Step 6 (Trigger System) — merged via PR feat(triggers): Trigger System (Phase 2 Step 6) #11
🟢 Step 7 (Shift Report) — THIS PR

After merge: Phase 2 COMPLETE. The blueprint's three-step plan delivered. Final write-up will live in docs/PHASE2-COMPLETE.md (similar to docs/PHASE1-COMPLETE.md) — I'll do that as a follow-up commit on main once you've burned in the shift report.

🤖 Generated with Claude Code

End-of-day summary of everything CODEC observed and accomplished. Single notification with type="shift_report" and a 5-section markdown body the PWA renders inline. Three trigger paths (time / idle / manual) with per-day dedup. 20 tests, all mocking filesystem (NO real ~/.codec writes). Components: codec_audit.py (+25 LOC) - 2 new event constants: SHIFT_REPORT_STARTED, SHIFT_REPORT_COMPLETED - PHASE2_STEP7_EVENTS frozenset - SHIFT_REPORT_EXTRA_FIELDS doc tuple skills/shift_report.py NEW (~470 LOC) - SKILL_NAME="shift_report", 8 trigger phrases, MCP-exposed - Manual entry: run(task) — also handles "[trigger=time]" / "[trigger=idle]" markers from observer - Internal entry: run_with_trigger_kind(kind) — used by observer - 5 section renderers (completed / blocked / observed / pending / tomorrow), each with a "no data" fallback for empty days - Inputs: audit.log + rotated logs (24h lookback configurable), notifications.json, observer_summaries/, skill_proposals/<today>/ - Per-day dedup via ~/.codec/shift_report_state.json (atomic write) - Optional auto-save to ~/Documents/CODEC Shift Reports/YYYY-MM-DD.md (config.shift_report.auto_save_path; default null = notification only) - Kill switch: SHIFT_REPORT_ENABLED env var (default true) codec_observer.py (+50 LOC) - importlib.util added at top - run_daemon() loop calls _maybe_fire_shift_report(idle) after each poll's long-idle reset check; try/except wraps so shift_report failures NEVER break observer polling - _maybe_fire_shift_report() lazy-imports shift_report module, checks kill switch + per-day dedup + config flag, then matches: time path: now.hour == daily_at_hour AND now.minute == daily_at_minute idle path: idle_seconds >= idle_minutes * 60 Whichever matches first fires; per-day dedup suppresses the other. tests/test_shift_report.py NEW (~440 LOC, 20 tests) Assembly (8): empty inputs / completed / blocked / observed_patterns / pending_decisions / proposals / tomorrow / observer_summaries Notification + state (3): post type=shift_report / dedup / manual bypass Kill switch + config (3): disabled / default / config overrides Audit emit (3): started+completed share cid / extras / kill-switch silent Observer integration (3): idle fires / off-window doesn't / dedup honored Test isolation: every storage path (audit_log, notifications.json, observer_summaries/, skill_proposals/, shift_report_state.json, config.json) redirected to tmp_path. NO real ~/.codec/* writes. NO Apple Reminders / Calendar / Notes. Audit emits redirected. AGENTS.md (+45 LOC) - §3 new "Shift Report (Phase 2 Step 7)" sub-section - §6 new audit events table - §10 new don't-touch entries (shift_report_state.json, SHIFT_REPORT_ENABLED env var, config.shift_report.{...} tunables) Test result: 20 passed in 0.17s. Side-effects clean: pending_questions: 0 Apple Reminders: 0 /tmp/codec_*.txt: 0 shift_report_state.json: absent (not written by tests) Diff inventory: ~+1,030 LOC functional + tests + docs. Phase 2 status after merge: 3/3 (Steps 5 + 6 + 7 all done). Phase 2 COMPLETE pending this PR + sign-off. Safety summary (what this PR explicitly does NOT do): - Does NOT auto-fire at merge time (observer's _maybe_fire_shift_report only fires when conditions match; with default config that's 18:00 local OR 30 min idle) - Does NOT touch _HTTP_BLOCKED - Does NOT create Apple Reminders / Notes / Calendar entries - Does NOT add a new PM2 service - Does NOT modify codec_self_improve.py (Step 4 work, sealed)

…(Steps 5/6/7) Phase 2 (Observer + Triggers + Shift Report) merged and production-stable: - Step 5 (PR #9 824a52f) + hotfix PR #10 (26e6add) — `observer.ocr_enabled` flag - Step 6 (PR #11 2d2ff3f) — Trigger System (matcher + cooldown + consent) - Step 7 (PR #12 0e40687) — end-of-day shift report Net: +91 passing tests (823/20/73), 0 new failures, 0 new skips. Live audit proof captured: shift_report_started+_completed paired emits at 2026-05-02T18:49:40Z with shared cid=5f188e5485e5.

AVADSA25 merged commit 0e40687 into main May 2, 2026
1 check passed

AVADSA25 mentioned this pull request May 2, 2026

hotfix: add shift_report to CHAT_SKILL_ALLOWLIST #13

Merged

3 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(shift_report): end-of-day shift report (Phase 2 Step 7)#12

feat(shift_report): end-of-day shift report (Phase 2 Step 7)#12
AVADSA25 merged 1 commit intomainfrom
phase2-step7-shift-report

AVADSA25 commented May 2, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

AVADSA25 commented May 2, 2026

Summary

Three trigger paths

5 sections, ~500-1500 words

Commits

Test plan

Test breakdown (20)

§6 audit events (2 new)

What this PR does NOT do

Kill switch

After merge

Phase 2 status after this PR

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants