Apple Flow

Prompt bundle for Apple Shortcuts-based automation flows.

Automation Prompt Bundle

Production-ready automation prompt library for Apple Shortcuts + Apple Flow, focused on codebase upkeep and agent-office upkeep.

Quick Start

  • Use explicit prefixes for deterministic behavior: idea:, plan:, task:, project:.
  • Use !!agent in Calendar/Reminders/Notes prompts.
  • Prefer plan: for recurring automations to reduce approval fatigue.
  • Use task: only when you intentionally want a mutating run and approval request.

Prompt Card Schema

Each card uses this exact schema:

  • Name
  • Goal
  • Best Channel (text, calendar, reminder, note)
  • Frequency
  • Flow Command (exact payload to send)
  • Why This Prefix
  • Response Contract
  • Approval Expectation
  • Failure Fallback

Output Style Invariants (apply to all cards)

  • Decision-first output ordering: recommendation, evidence, next action.
  • Mobile readable: 4-8 bullets in normal mode.
  • Single-screen summary first, details second.
  • Emoji section headers in every recurring prompt.

A) Repo Upkeep Automations (24)

Pack A1: Daily Repo Pulse (6)

R01

  • Name: Morning Health Snapshot
  • Goal: Detect operational problems early.
  • Best Channel: calendar
  • Frequency: Weekdays 08:15
  • Flow Command:
!!agent plan: Produce an engineering morning snapshot for this repo.
Output sections:
🏥 Health
📊 Usage Today
⏳ Pending Approvals
🚨 Immediate Risks
➡️ First Action
Max length: 9 bullets total.
  • Why This Prefix: plan: gives decisioning without file mutation.
  • Response Contract: Recommendation in first bullet under ➡️ First Action, then evidence.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R02

  • Name: Today Maintenance Queue
  • Goal: Convert recent failures/history into a ranked maintenance queue.
  • Best Channel: text
  • Frequency: Daily 09:00
  • Flow Command:
plan: Build today's maintenance queue from recent errors, approvals, and history.
Output sections:
🧭 Queue Priority (P0/P1/P2)
🧪 Risk if Delayed
📌 Suggested Owner
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: planning-only backlog synthesis.
  • Response Contract: Include exactly top 5 items with impact/effort confidence.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R03

  • Name: Top 3 Risky Files
  • Goal: Surface highest-risk files by churn + failures.
  • Best Channel: calendar
  • Frequency: Weekdays 11:30
  • Flow Command:
!!agent plan: Identify top 3 risky files from recent churn and failing tests.
Output sections:
🔥 Top 3 Files
🧾 Why Risky
🛠️ Guardrail to Add
➡️ First Action
Max length: 7 bullets.
  • Why This Prefix: non-mutating risk analysis.
  • Response Contract: Include one test or check recommendation per risky file.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R04

  • Name: Branch/PR Hygiene Summary
  • Goal: Keep branches/PR workflow tidy.
  • Best Channel: text
  • Frequency: Daily 14:00
  • Flow Command:
plan: Summarize branch and PR hygiene.
Output sections:
🌿 Branch State
🔀 PR Backlog
🧼 Hygiene Gaps
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: status + recommendations without mutating.
  • Response Contract: Flag stale branches/PRs with suggested close/merge/rebase action.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R05

  • Name: Dependency Drift Watch
  • Goal: Spot upgrade and security drift.
  • Best Channel: calendar
  • Frequency: Weekdays 16:00
  • Flow Command:
!!agent plan: Assess dependency drift and upgrade risk.
Output sections:
📦 Drift Summary
🛡️ Security-Relevant Packages
⚖️ Upgrade Risk
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: non-mutating drift plan.
  • Response Contract: Distinguish urgent security drift vs routine drift.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R06

  • Name: End-of-Day Engineering Recap
  • Goal: Capture what moved and what's next.
  • Best Channel: calendar
  • Frequency: Weekdays 18:00
  • Flow Command:
!!agent plan: Draft end-of-day engineering recap.
Output sections:
✅ Completed
⚠️ Blockers
📈 Quality Signals
➡️ First Action Tomorrow
Max length: 9 bullets.
  • Why This Prefix: recap drafting without edits.
  • Response Contract: Include one carry-forward priority for tomorrow.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

Pack A2: Code Quality & Test Reliability (6)

R07

  • Name: Flaky Test Triage
  • Goal: Isolate flaky tests and propose stabilizing strategy.
  • Best Channel: text
  • Frequency: Daily 10:30
  • Flow Command:
plan: Triage flaky tests and propose stabilization sequence.
Output sections:
🧪 Suspected Flakes
🧠 Root-Cause Hypothesis
🛠️ Stabilization Plan
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: diagnosis and sequence planning.
  • Response Contract: Rank flakes by impact to release confidence.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R08

  • Name: Critical Path Coverage Gaps
  • Goal: Find critical-path areas with weak test coverage.
  • Best Channel: calendar
  • Frequency: Tue/Thu 13:00
  • Flow Command:
!!agent plan: Find uncovered critical paths and propose minimum high-value tests.
Output sections:
🎯 Critical Paths
📉 Coverage Gaps
🧪 Minimum Tests to Add
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: analysis-first, low approval burden.
  • Response Contract: Recommend up to 3 tests with ROI rationale.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R09

  • Name: Smallest Safe Refactor Candidates
  • Goal: Identify low-risk refactors with high maintainability gain.
  • Best Channel: text
  • Frequency: Daily 15:00
  • Flow Command:
plan: Identify smallest safe refactors to reduce complexity and risk.
Output sections:
🧩 Candidate Refactors
📏 Estimated Scope
🧪 Verification Needed
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: keeps output tactical, non-mutating.
  • Response Contract: Include one "do now" and one "defer" recommendation.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R10

  • Name: Lint/Type Noise Reduction
  • Goal: Reduce alert fatigue from lint/type noise.
  • Best Channel: reminder
  • Frequency: Daily 16:30
  • Flow Command:
!!agent plan: Propose lint/type noise reduction without weakening guardrails.
Output sections:
🔎 Noise Sources
🛡️ Guardrails to Keep
🧹 Cleanup Plan
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: design-level recommendations first.
  • Response Contract: Separate policy/config cleanup from true bug risks.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R11

  • Name: Repro Checklist Generator
  • Goal: Turn failing traces into reproducible checklists.
  • Best Channel: note
  • Frequency: On-demand
  • Flow Command:
!!agent plan: Generate a reproducibility checklist for recent failing traces.
Output sections:
🧾 Repro Preconditions
▶️ Step Sequence
✅ Expected vs Actual
🧪 Isolation Tests
➡️ First Action
Max length: 10 bullets.
  • Why This Prefix: artifact drafting without writing repo files.
  • Response Contract: Include env assumptions and confidence level.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R12

  • Name: Regression Prevention Checklist
  • Goal: Build guardrails around recently touched modules.
  • Best Channel: calendar
  • Frequency: Weekdays 17:00
  • Flow Command:
!!agent plan: Build regression-prevention checklist for recently touched modules.
Output sections:
📦 Touched Modules
⚠️ Regression Vectors
🛡️ Preventive Checks
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: no mutation, explicit guardrail design.
  • Response Contract: Include fast pre-merge and slow nightly checks.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

Pack A3: Issue/PR Workflow (6)

R13

  • Name: Open Issue Impact/Effort Triage
  • Goal: Prioritize issue backlog.
  • Best Channel: text
  • Frequency: Daily 12:00
  • Flow Command:
plan: Triage open issues by impact, effort, and urgency.
Output sections:
📋 Top Queue
⚖️ Impact/Effort Notes
🚫 Items to Defer
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: structured triage without actions.
  • Response Contract: Use P0/P1/P2 tags and confidence labels.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R14

  • Name: Stale Issue Close-or-Advance
  • Goal: Decide stale issue fate.
  • Best Channel: calendar
  • Frequency: Mon/Wed/Fri 10:00
  • Flow Command:
!!agent plan: For stale issues, decide close, defer, or advance with criteria.
Output sections:
🕰️ Stale Candidates
✅ Close Rationale
🚀 Advance Plan
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: policy-based recommendation.
  • Response Contract: Include exact close message template suggestion.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R15

  • Name: Release Readiness Gate Review
  • Goal: Assess release blockers and go/no-go.
  • Best Channel: calendar
  • Frequency: Tue/Thu 18:30
  • Flow Command:
!!agent plan: Perform release-readiness gate review.
Output sections:
🚦 Go/No-Go Recommendation
🧪 Test Gate
🛡️ Risk Gate
📚 Docs/Comms Gate
➡️ First Action
Max length: 9 bullets.
  • Why This Prefix: decision memo style, no file edits.
  • Response Contract: Start with one-line go/no-go decision.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R16

  • Name: Changelog Draft from Recent Work
  • Goal: Prepare human-readable release notes.
  • Best Channel: text
  • Frequency: Daily 17:30
  • Flow Command:
plan: Draft today's changelog from recent work.
Output sections:
✨ User-Facing Changes
🧰 Internal Improvements
🐛 Fixes
⚠️ Breaking/Behavioral Notes
➡️ First Action
Max length: 10 bullets.
  • Why This Prefix: content drafting without committing changes.
  • Response Contract: Keep each bullet release-note ready.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R17

  • Name: Reviewer-Facing Summary Generator
  • Goal: Reduce reviewer time-to-context.
  • Best Channel: reminder
  • Frequency: Daily 14:30
  • Flow Command:
!!agent plan: Create reviewer-facing summary for active PRs.
Output sections:
🧭 What Changed
⚠️ Risk Areas
🧪 How to Validate
❓ Open Questions
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: non-mutating summary generation.
  • Response Contract: Include explicit reviewer checklist.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R18

  • Name: Post-Merge Verification Checklist
  • Goal: Confirm merged changes are healthy.
  • Best Channel: calendar
  • Frequency: Daily 19:00
  • Flow Command:
!!agent plan: Produce post-merge verification checklist for today's merged work.
Output sections:
✅ Smoke Checks
📈 Observability Checks
🚨 Rollback Triggers
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: operational checklist drafting.
  • Response Contract: Include pass/fail criteria for each check.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

Pack A4: Security/Ops Hygiene (6)

R19

  • Name: Security Best-Practice Review
  • Goal: Regularly surface security hardening opportunities.
  • Best Channel: calendar
  • Frequency: Weekdays 09:30
  • Flow Command:
!!agent plan: Run a security best-practice review for current repo changes.
Output sections:
🛡️ Key Findings
🔥 Priority Risks
🧰 Quick Mitigations
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: review mode without applying fixes.
  • Response Contract: Prioritize findings by severity and exploitability.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R20

  • Name: Secrets/Config Exposure Sanity Check
  • Goal: Catch accidental secret leakage and risky config drift.
  • Best Channel: text
  • Frequency: Daily 13:30
  • Flow Command:
plan: Check for potential secrets/config exposure and risky defaults.
Output sections:
🔐 Exposure Signals
⚙️ Config Drift
🧯 Containment Steps
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: analysis and mitigation sequencing only.
  • Response Contract: Distinguish probable leak vs false positive.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R21

  • Name: Approval Backlog Aging Monitor
  • Goal: Keep approval queue healthy.
  • Best Channel: calendar
  • Frequency: Every weekday at 12:15 and 17:15
  • Flow Command:
!!agent plan: Monitor approval backlog aging and propose triage actions.
Output sections:
⏳ Aged Approvals
⚠️ Expiring Soon
🧭 Triage Recommendation
➡️ First Action
Max length: 7 bullets.
  • Why This Prefix: no mutation, queue policy recommendations.
  • Response Contract: Include threshold-based intervention advice.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R22

  • Name: Alert/Runbook Drift Detector
  • Goal: Keep runbooks aligned with current system behavior.
  • Best Channel: note
  • Frequency: Weekly Monday 11:00
  • Flow Command:
!!agent plan: Detect drift between observed alerts and current runbooks.
Output sections:
🔔 Alert Patterns
📘 Runbook Gaps
🛠️ Suggested Updates
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: drift assessment and proposal only.
  • Response Contract: Include one minimal runbook patch suggestion.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R23

  • Name: Next Ops Automation Candidate
  • Goal: Identify highest ROI operational automation.
  • Best Channel: text
  • Frequency: Weekly Tuesday 09:45
  • Flow Command:
idea: What is the highest-ROI ops task to automate next in this repo workflow?
Output sections:
💡 Candidate Automation
📈 ROI Rationale
🧱 Dependencies
➡️ First Action
Max length: 7 bullets.
  • Why This Prefix: exploratory ideation, no execution.
  • Response Contract: Recommend one candidate and one fallback candidate.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

R24

  • Name: Incident Follow-Up Synthesis
  • Goal: Turn incidents into durable corrective actions.
  • Best Channel: note
  • Frequency: On-demand after incidents
  • Flow Command:
plan: Synthesize latest incident learnings into corrective-action plan.
Output sections:
🧠 Root Cause Summary
🧯 Immediate Fixes
🛡️ Prevention Actions
📏 Verification Criteria
➡️ First Action
Max length: 10 bullets.
  • Why This Prefix: postmortem planning, no immediate file changes.
  • Response Contract: Include owner suggestion and time horizon per action.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

B) Agent-Office Upkeep Automations (24)

Pack B1: Inbox + Daily Operations (8)

O01

  • Name: Inbox Triage Sweep
  • Goal: Keep 00_inbox/inbox.md processed and actionable.
  • Best Channel: calendar
  • Frequency: Weekdays 08:40
  • Flow Command:
!!agent plan: Triage agent-office inbox entries into Keep, Do, Delegate, Archive candidates.
Output sections:
📥 Keep
⚙️ Do
🤝 Delegate
🗄️ Archive Candidates
➡️ First Action
Max length: 9 bullets.
  • Why This Prefix: categorization without mutating files.
  • Response Contract: Respect append-only inbox and summarize-before-archive rule.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O02

  • Name: Daily Note Bootstrap
  • Goal: Prepare 10_daily/YYYY-MM-DD.md agenda.
  • Best Channel: calendar
  • Frequency: Weekdays 07:45
  • Flow Command:
!!agent plan: Draft daily note bootstrap with top 3 priorities and open loops.
Output sections:
🎯 Top 3 Priorities
📅 Calendar & Commitments
🔁 Open Loops
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: output draft first, no file write.
  • Response Contract: Priorities must be concrete and time-bound.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O03

  • Name: Open Loops Extractor
  • Goal: Consolidate unresolved commitments.
  • Best Channel: text
  • Frequency: Daily 12:30
  • Flow Command:
plan: Extract open loops from inbox, reminders, and calendar context.
Output sections:
🔁 Open Loops
⏱️ Deadline Pressure
🧩 Dependencies
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: synthesis and prioritization only.
  • Response Contract: Include top 5 loops max with urgency tags.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O04

  • Name: Archive Candidates with Rationale
  • Goal: Identify what can be archived safely.
  • Best Channel: reminder
  • Frequency: Weekdays 17:40
  • Flow Command:
!!agent plan: Recommend what can be archived today with summary-first rationale.
Output sections:
🗄️ Archive Candidates
📝 Required Summary Before Archive
⚠️ Keep Active
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: policy-safe archive recommendations.
  • Response Contract: Never recommend deletion; archive only after summary.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O05

  • Name: End-of-Day Reflection Draft
  • Goal: Build concise reflection and carry-forward list.
  • Best Channel: calendar
  • Frequency: Weekdays 19:15
  • Flow Command:
!!agent plan: Draft end-of-day reflection for agent-office daily note.
Output sections:
✅ Wins
⛔ Blockers
➡️ Carry Forward
Max length: 7 bullets.
  • Why This Prefix: draft reflection without file mutation.
  • Response Contract: Include at least one blocker mitigation suggestion.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O06

  • Name: Next-Day Prep Checklist
  • Goal: Pre-stage tomorrow's focus.
  • Best Channel: calendar
  • Frequency: Weekdays 20:00
  • Flow Command:
!!agent plan: Prepare next-day checklist from current open loops and calendar.
Output sections:
📋 Prep Checklist
🧠 Context to Preserve
🚫 Risks to Avoid
➡️ First Action Tomorrow
Max length: 8 bullets.
  • Why This Prefix: planning and checklist generation only.
  • Response Contract: Include first 30-minute startup action.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O07

  • Name: Untriaged Inbox Nudge
  • Goal: Prevent inbox stagnation.
  • Best Channel: calendar
  • Frequency: Weekdays 11:45
  • Flow Command:
!!agent plan: Nudge me on untriaged inbox items and suggest 10-minute triage move.
Output sections:
📥 Untriaged Count
⚠️ Oldest Items
⏲️ 10-Minute Triage Sprint
➡️ First Action
Max length: 6 bullets.
  • Why This Prefix: low-friction nudge; no changes.
  • Response Contract: Keep short and action-biased.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O08

  • Name: Context Handoff Summary
  • Goal: Create handoff summary for interrupted days.
  • Best Channel: text
  • Frequency: Daily 18:45
  • Flow Command:
plan: Generate context handoff summary for tomorrow or delegate.
Output sections:
🧭 Current State
🔁 In-Flight Items
📌 Next Decision Needed
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: handoff drafting without mutation.
  • Response Contract: Include one "resume here" line.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

Pack B2: Memory Hygiene (8)

O09

  • Name: MEMORY.md Delta Proposal
  • Goal: Keep durable memory current and factual.
  • Best Channel: note
  • Frequency: Weekly Monday 08:30
  • Flow Command:
!!agent plan: Propose factual delta updates for agent-office/MEMORY.md from recent activity.
Output sections:
🧠 Add
✏️ Refine
🗑️ Remove or De-emphasize
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: change proposal only.
  • Response Contract: No speculation; factual compact statements.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O10

  • Name: Topic Memory Update Recs
  • Goal: Keep 60_memory/*.md useful and fresh.
  • Best Channel: calendar
  • Frequency: Weekly Wednesday 09:15
  • Flow Command:
!!agent plan: Recommend updates for topic memory files in 60_memory.
Output sections:
📂 Topics to Update
🧾 Proposed Facts
🔗 Cross-Topic Links
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: recommendation mode only.
  • Response Contract: One bullet per topic, max five topics.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O11

  • Name: Memory Contradiction Detector
  • Goal: Identify conflicting facts/preferences.
  • Best Channel: reminder
  • Frequency: Weekly Friday 10:45
  • Flow Command:
!!agent plan: Detect contradictions across MEMORY.md and 60_memory files.
Output sections:
⚠️ Contradictions
🔎 Evidence
🧭 Resolution Suggestion
➡️ First Action
Max length: 7 bullets.
  • Why This Prefix: analytical QA, no edits.
  • Response Contract: Include conflict confidence for each contradiction.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O12

  • Name: Memory Staleness Alert
  • Goal: Highlight stale sections that need refresh.
  • Best Channel: calendar
  • Frequency: Weekly Thursday 16:15
  • Flow Command:
!!agent plan: Identify stale memory sections and propose refresh order.
Output sections:
🕰️ Stale Sections
📈 Impact of Staleness
🛠️ Refresh Order
➡️ First Action
Max length: 7 bullets.
  • Why This Prefix: non-mutating hygiene prioritization.
  • Response Contract: Prioritize by decision impact, not file age alone.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O13

  • Name: Preference Extraction from Decisions
  • Goal: Convert repeated decisions into explicit preferences.
  • Best Channel: text
  • Frequency: Weekly Tuesday 15:30
  • Flow Command:
idea: Extract likely stable preferences from recent decisions for memory hardening.
Output sections:
🧭 Candidate Preferences
🧾 Evidence
✅ Confidence
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: exploratory inference before acceptance.
  • Response Contract: Mark each preference as confirmed or needs validation.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O14

  • Name: Guardrail Drift Checker
  • Goal: Verify behavior against SCAFFOLD non-negotiables.
  • Best Channel: calendar
  • Frequency: Weekly Monday 14:00
  • Flow Command:
!!agent plan: Check drift against agent-office/SCAFFOLD.md non-negotiables.
Output sections:
🛡️ Guardrails Compliant
⚠️ Guardrails at Risk
🧯 Corrective Suggestions
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: compliance review without mutation.
  • Response Contract: Explicitly call out append-only and summarize-before-archive compliance.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O15

  • Name: Weekly Memory Compacting Plan
  • Goal: Reduce noise and redundancy in memory system.
  • Best Channel: note
  • Frequency: Weekly Sunday 17:00
  • Flow Command:
!!agent plan: Create weekly memory compacting plan for MEMORY.md and 60_memory.
Output sections:
🧹 Redundancy Hotspots
📦 Merge Candidates
🚫 Do-Not-Compact Items
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: planning mode for controlled consolidation.
  • Response Contract: Preserve critical context and confidence caveats.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O16

  • Name: Promote-or-Prune Candidates
  • Goal: Decide what to promote to durable memory or prune from active focus.
  • Best Channel: calendar
  • Frequency: Weekly Friday 18:10
  • Flow Command:
!!agent plan: Recommend promote-or-prune candidates from current notes/memory.
Output sections:
📈 Promote
🪶 Keep as Ephemeral
🗂️ Prune/Archive Candidate
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: curation without direct file edits.
  • Response Contract: Tie each recommendation to expected utility.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

Pack B3: Project Brief Governance (8)

O17

  • Name: Project Brief Status Normalizer
  • Goal: Keep 20_projects/*/brief.md statuses coherent.
  • Best Channel: note
  • Frequency: Weekly Monday 10:30
  • Flow Command:
!!agent plan: Normalize project brief statuses and suggest status corrections.
Output sections:
📂 Projects Reviewed
🔄 Status Corrections
📌 Missing Brief Fields
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: governance recommendations only.
  • Response Contract: Require evidence for each status correction.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O18

  • Name: Milestone Aging Checker
  • Goal: Catch slipping milestones early.
  • Best Channel: calendar
  • Frequency: Tue/Thu 11:15
  • Flow Command:
!!agent plan: Check milestone aging across active project briefs.
Output sections:
⏳ Aging Milestones
⚠️ Slippage Risk
🧭 Recovery Option
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: timeline risk analysis only.
  • Response Contract: Include one fast recovery move per high-risk milestone.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O19

  • Name: Risk Register Refresh
  • Goal: Keep project risk register current.
  • Best Channel: reminder
  • Frequency: Weekly Wednesday 17:20
  • Flow Command:
!!agent plan: Refresh risk register suggestions for active projects.
Output sections:
🔥 New Risks
🛡️ Mitigations
📏 Trigger Conditions
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: planning and risk framing.
  • Response Contract: Provide probability/impact hints for each risk.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O20

  • Name: Blocked Project Escalation Summary
  • Goal: Escalate blocked work with clarity.
  • Best Channel: text
  • Frequency: Daily 16:50
  • Flow Command:
plan: Summarize blocked projects and produce escalation-ready notes.
Output sections:
⛔ Blocked Projects
🧱 Blocking Dependency
📣 Escalation Draft
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: communication planning without mutation.
  • Response Contract: Include one escalation message template.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O21

  • Name: Priority Rebalance Across Projects
  • Goal: Reallocate attention based on impact and deadlines.
  • Best Channel: calendar
  • Frequency: Weekly Monday 13:30
  • Flow Command:
!!agent plan: Rebalance priorities across active projects.
Output sections:
🎯 Priority Order
⚖️ Tradeoff Notes
🧩 Capacity Fit
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: strategic planning output.
  • Response Contract: Include what to pause to protect top priority.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O22

  • Name: Weekly Portfolio Snapshot
  • Goal: Produce one-screen project portfolio health view.
  • Best Channel: calendar
  • Frequency: Friday 15:45
  • Flow Command:
!!agent plan: Generate weekly portfolio snapshot for active projects.
Output sections:
🟢 Green
🟡 Yellow
🔴 Red
➡️ First Action
Max length: 9 bullets.
  • Why This Prefix: reporting/synthesis only.
  • Response Contract: Exactly one priority action per color band.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O23

  • Name: Completion/Archive Recommendation
  • Goal: Identify projects ready to close/archive correctly.
  • Best Channel: reminder
  • Frequency: Weekly Thursday 18:30
  • Flow Command:
!!agent plan: Recommend project completion/archive candidates requiring summary first.
Output sections:
✅ Ready to Complete
📝 Required Completion Summary
🗄️ Archive Candidate
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: governance decision support.
  • Response Contract: Never skip summary-before-archive requirement.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

O24

  • Name: Next-Action Generator per Brief
  • Goal: Convert each active brief into immediate next action.
  • Best Channel: text
  • Frequency: Daily 09:20
  • Flow Command:
plan: For each active project brief, generate one concrete next action.
Output sections:
📂 Project
➡️ Next Action
📏 Definition of Done
⏱️ Estimated Effort
Max length: 8 bullets.
  • Why This Prefix: tactical planning with no file mutation.
  • Response Contract: One action per project, max 5 projects.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

C) Meta-Automation / Flow Control (12)

Pack C1: Mute/Unmute Guardrails (2)

M01

  • Name: Focus Window Mute Guard
  • Goal: Automatically enforce focus quieting behavior.
  • Best Channel: calendar
  • Frequency: Weekdays 09:55 and 13:55
  • Flow Command:
!!agent system: mute
  • Why This Prefix: direct control command.
  • Response Contract: Confirm mute state and next unmute checkpoint.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

M02

  • Name: Focus Window Unmute Guard
  • Goal: Resume proactive companion messages after focus blocks.
  • Best Channel: calendar
  • Frequency: Weekdays 12:05 and 17:05
  • Flow Command:
!!agent system: unmute
  • Why This Prefix: direct control command.
  • Response Contract: Confirm unmuted and summarize missed critical events if available.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

Pack C2: Approval Backlog Interventions (2)

M03

  • Name: Approval Backlog Sweep
  • Goal: Keep pending approvals actionable.
  • Best Channel: calendar
  • Frequency: Weekdays 12:10
  • Flow Command:
!!agent status
  • Why This Prefix: built-in command for pending approvals.
  • Response Contract: Return concise list with aging + suggested triage order.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

M04

  • Name: Approval Intervention Advisor
  • Goal: Decide what to approve now vs deny/defer.
  • Best Channel: text
  • Frequency: Weekdays 12:20
  • Flow Command:
plan: Based on pending approvals, recommend approve/deny/defer with rationale.
Output sections:
✅ Approve Now
🚫 Deny
⏳ Defer
➡️ First Action
Max length: 7 bullets.
  • Why This Prefix: decision support without executing approvals.
  • Response Contract: Include risk of delay and confidence tags.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

Pack C3: Automation Effectiveness Review (2)

M05

  • Name: Weekly Automation Keep/Remove/Tune
  • Goal: Keep automation set high-signal.
  • Best Channel: calendar
  • Frequency: Sunday 18:30
  • Flow Command:
!!agent plan: Review automation effectiveness and classify each as keep/remove/tune.
Output sections:
✅ Keep
🗑️ Remove
🎛️ Tune
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: policy review and tuning guidance.
  • Response Contract: Include reason and expected impact per recommendation.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

M06

  • Name: Noise Budget Audit
  • Goal: Prevent automation spam.
  • Best Channel: reminder
  • Frequency: Weekly Friday 17:50
  • Flow Command:
!!agent plan: Audit message noise budget and propose cadence reductions where needed.
Output sections:
📣 High-Noise Automations
📉 Reduction Suggestions
🧪 Measurement Plan
➡️ First Action
Max length: 7 bullets.
  • Why This Prefix: non-mutating optimization.
  • Response Contract: Quantify expected signal/noise improvement.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

Pack C4: Prompt Quality Auditor (2)

M07

  • Name: Prompt Quality Scorecard
  • Goal: Score prompts for clarity and execution reliability.
  • Best Channel: note
  • Frequency: Weekly Tuesday 18:00
  • Flow Command:
!!agent plan: Audit current automation prompts for verbosity, specificity, and ambiguity.
Output sections:
🧪 Scorecard
✂️ Simplify
🎯 Clarify
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: quality analysis only.
  • Response Contract: Score each dimension from 1-5 with one fix per low score.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

M08

  • Name: Prefix Intent Auditor
  • Goal: Ensure prefixes match intent and approval profile.
  • Best Channel: text
  • Frequency: Weekly Wednesday 12:40
  • Flow Command:
plan: Audit whether prompt prefixes match intent (idea/plan/task/project).
Output sections:
🧭 Prefix Mismatches
🛡️ Approval Risk
🔄 Recommended Prefix Changes
➡️ First Action
Max length: 7 bullets.
  • Why This Prefix: classification audit only.
  • Response Contract: Highlight high-risk task: misuses.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

Pack C5: Cost/Usage Governance (2)

M09

  • Name: Daily Usage Pulse
  • Goal: Monitor day-level consumption.
  • Best Channel: calendar
  • Frequency: Weekdays 16:10
  • Flow Command:
!!agent usage today
  • Why This Prefix: built-in usage command.
  • Response Contract: Include quick interpretation: on-track or elevated.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

M10

  • Name: Monthly Usage Governance Review
  • Goal: Control cost trends and block hotspots.
  • Best Channel: calendar
  • Frequency: 1st of month 09:10
  • Flow Command:
!!agent plan: Analyze monthly usage and block-level hotspots; suggest cost controls.
Output sections:
💰 Monthly Trend
🧱 Cost Hotspots
🎛️ Control Levers
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: analytical governance, no mutation.
  • Response Contract: Include one low-risk cost optimization.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

Pack C6: Trigger Integrity Check (2)

M11

  • Name: Trigger Tag Coverage Check
  • Goal: Ensure recurring automations include !!agent where required.
  • Best Channel: calendar
  • Frequency: Weekly Monday 09:05
  • Flow Command:
!!agent plan: Check trigger integrity for calendar/reminder/note automations and tag coverage.
Output sections:
🏷️ Missing Tag Risks
✅ Valid Triggers
🛠️ Repair Suggestions
➡️ First Action
Max length: 7 bullets.
  • Why This Prefix: integrity audit without changes.
  • Response Contract: Flag likely silent-failure automations first.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

M12

  • Name: Cadence Conflict Checker
  • Goal: Detect overlapping schedule conflicts and collision windows.
  • Best Channel: note
  • Frequency: Weekly Sunday 19:00
  • Flow Command:
!!agent plan: Detect cadence conflicts across automations and suggest conflict-free schedule.
Output sections:
🕒 Collision Windows
⚠️ Potential Redundancy
🎯 Proposed Cadence
➡️ First Action
Max length: 8 bullets.
  • Why This Prefix: schedule optimization analysis.
  • Response Contract: Provide one simplified weekly cadence map.
  • Approval Expectation: No
  • Failure Fallback: If data unavailable, report No signal and list what was missing.

Rollout Activation Plan

Phase 1 (Pilot: week 1, 12 prompts)

  • Repo: R01, R02, R03, R07, R15, R21
  • Agent-office: O01, O02, O03, O09, O14, O24

Target outcomes:

  • Detect drift earlier.
  • Keep approval queue manageable.
  • Improve daily focus and context continuity.

Phase 2 (Expansion: week 2)

  • Add remaining Repo and Agent-office prompts (R04-R24, O04-O23).
  • Tune noisy prompts by adjusting cadence and output max bullets.

Phase 3 (Optimization: week 3+)

  • Activate all Meta prompts (M01-M12).
  • Run monthly prune of low-value automations.

Validation Checklist

  1. Classification correctness
  • Verify each prompt lands in expected mode (idea/plan/task/project or control command).
  1. Approval safety
  • Verify any intentional mutating prompt triggers approval and is sender-attributable.
  1. Channel trigger validity
  • Verify !!agent-tagged calendar/reminder/note prompts are ingested and untagged ones are ignored.
  1. Signal quality
  • Verify each prompt returns actionable structured output under sparse data.
  1. Noise control
  • Verify no redundant overlapping proactive messages in daily windows.
  1. Agent-office integrity
  • Verify inbox append-only and summarize-before-archive behavior are preserved.
  1. Readability acceptance
  • Verify iMessage outputs remain concise with emoji headers and scannable bullets.